Added two new modules, the MOM6-specific MOM_check_scaling.F90 and the generic
framework module MOM_scaling_check.F90, to assess the uniqueness of the unit
scaling factors for all of the variables used by MOM6.  If there are overlaps in
scaling factors for different units, this also identifies and suggests alternate
scaling factors with less overlaps.  This commit includes the introduction of
the new publicly visible routines check_scaling_factors(), scales_to_powers()
and check_MOM6_scaling_factors.  This new capability does not do anything for
sufficiently low levels of model verbosity, and it is silent if the scaling
factors are unique, or if less than 2 dimensions are being rescaled.  All
answers and output files are bitwise identical, but there can be additional
messages to stdout.

  Renamed the framework module MOM_scaling_check.F90 to MOM_unique_scales.F90 to
help differentiate it from MOM_check_scaling.F90, and renamed the subroutine
check_scaling_factors() as check_scaling_uniqueness().  Also added
_Dimensional_consistency.dox to describe the dimensional consistency testing.
This commit should address the issues raised in the review of MOM6 PR #49.  All
answers and output are bitwise identical.

  Corrected memory declarations in MOM_regridding.F90 in cases where the
vertical size of the input and output grids are not the same.  Although this is
not know to have caused any particular problems, these inconsistencies could
lead to segmentation faults in cases where the target grid (e.g., diagnostic
output) is larger than the input grid (e.g., the model's native grid).  In some
cases, certain grid generation options were only written to work with the same
size of input and output grids, and error handling has been added to these cases
to gracefully bring down the model if they are used with different grid sizes.
All answers are bitwise identical in the MOM6-examples test suite, but it is
conceivable that this could correct subtle (memory-related) issues in some
configurations.

 - default behavior returns field size using fms1 format.

…65)

* Eliminate GET_ALL_PARAMS in hor_visc_init

  Added do_not_log arguments to get_param calls in MOM_hor_visc.F90 that are
only used conditionally, and eliminated the unlogged GET_ALL_PARAMS runtime
parameter and get_all variable in hor_visc_init().  By design, all logging of
parameters after this commit is identical to before, even for variables that are
inactive and therefore should not be logged.  In several places, there were some
problems, mostly with the GME code, that have been noted in comments marked with
'###'.  Also cleaned up the code alignment and eliminated unneeded temporary
variables in a few places in hor_visc().  All solutions are bitwise identical,
and no output is changed.

* Move call to get get_KH outside k-loop

The call to get the 3-d GME diffusivity arrays and the subsequent
blocking halo update was moved outside of the k-loop.

* Increase loop range for calculation of GME fluxes

* Makes GME filter rotationally invariant

* Makes the GME filter rotationally invariant
* Adds a runtime param to allow the user to control how many
smoothing passes should be done.

* Rearranges the get_param calls related to Leith

The get_param calls for Leith were not in the correct location.
This commit fixes that.

* Adding halo updates

* Fixes do loops indices and adds diagnostics

* Adds option to save barotropic tension and strain;

* Fixes many i and j loops indices associated with GME to avoid halo
problems and unnecessary halo updates. With these changes, the
model is now conserving mass and tracers when USE_GME = True.

* Fixes issues related to the merging with dev/gfdl

* Fixes calculation of FrictWork_GME

The calculation now mimics the calculation of FrictWork and it
includes the energy diffusion term.

* Removes dependency of FrictWork_GME calculation on MEKE

* Adding sh_xy_sq and sh_xx_sq in the OMP directives

Co-authored-by: Robert Hallberg <[email protected]>

Subroutine adjustEtaToFitBathymetry had a hard-wired parameter
(hTolerance = 0.1) controlling the tolerance when adjusting the
thickness to fit the bathymetry. This patch adds an user-controlled
parameter (THICKNESS_TOLERANCE), which replaces hTolerance.
THICKNESS_TOLERANCE is only activated when ADJUST_THICKNESS=True.

Main to dev

  Actually calculate the mean temperature and salinity reported by
MOM_state_stats().  Due to an oversight, these means were always being reported
as 0.  This changes the output when the debugging flag DEBUG_CONSERVATION=True.
All answers are bitwise identical.

  Added the new function global_mass_int_EFP(), which is analogous to
global_mass_integral but returns its result in extended fixed point (EFP_type)
format and always uses reproducing sums, to facilitate layout-invariant
global integrals but with the potential for deferred global reductions so that
this last step can be combined for various global reductions for efficiency.
All answers are bitwise identical, but there is a new public interface.

  Use global_mass_integral for the debugging diagnostics of the tracer amounts
before and after diffusion in lateral_boundary_diffusion, and replaced a call to
write(*,*) with a call to MOM_mesg to actually write the message.  The
global_mass_integral uses reproducing sums, and is invariant to layout, while
MOM_mesg is preferable for output because it will allow us to more cleanly
control how output is handled and which processors do the writing.  All
solutions are bitwise identical, although some debugging output will change.

  Use reproducing sums for tabulating tracer stocks, and move the global sum for
the tracer stocks form write_energy into call_tracer_stocks.  This involves
changes to the type of an argument (from real to EFP_type) for two arguments to
the internal routine store_stocks.  Existing tracer stock packages will still
work, but to benefit from the reproducing sums, they will also have to change
their reported values from real to EFP_type.  This is demonstrated for two
packages (advection_test_tracer and ideal_age_example), where the stocks are now
found with calls to global_mass_int_EFP(), replacing the previous explicit
sums.  With this change, the reported stock values from these packages are
identical for different PE layouts and can be much more accurate than before,
but they are different from the previously reported values at roundoff (for
positive-definite tracers), but it could be larger for tracers with a near-zero
mean value.  All solutions are bitwise identical, but output changes.

  Modified the remaining tracer packages to use the reproducing stocks.  The
reported stock values from these packages will have changed slightly, but they
now reproduce across PE layouts.  All solutions are bitwise identical, but
output changes.

  Removed trailing white space on two lines.  All answers are bitwise identical.

  Added new tmp_scale optional arguments to global_area_mean, global_layer_mean,
global_area_mean_u, global_area_mean_v, global_area_integral, global_volume_mean
and global_mass_integral.  This argument allows the reproducing sums to work
with rescaled variables without undoing the scaling of the returned variable.
Also corrected the area_rescaling in global_area_mean_u and global_area_mean_v,
which were substantially changing answers when horizontal distances were being
rescaled (i.e., if L_RESCALE_POWER is not 0).  These routines had only been
added recently, so this change is only changing answers with HOMOGENIZE_FORCINGS
= True.

  Add optional tmp_scale arguments to the various homogenize_field routines and
use them in numerous calls in homogenize_forcing and homogenize_mech_forcing, so
that the answers will pass the dimensional rescaling tests for large rescaling
factors when HOMOGENIZE_FORCINGS = True.  Among other changes is the addition of
a verticalGrid_type argument to homogenize_forcing.  All answers are bitwise
identical in the existing MOM6-examples test suite, but it will change answers
in other cases with dimensional rescaling active and homogenized forcing.

This patch includes several minor changes to the MOM_file_parser and
supporting modules in order to accommodate stronger unit testing.

It includes the following API changes:

- Removal of `static_value` from `get_param`

- Redefined `link_parameter` and `parameter_block` as private

- New functions: `all_across_PEs()`, `any_across_PEs()`

`static_value` was not used in any known experiments (outside of
internal GFDL testing), and the two derived types describe internal
operations within `MOM_file_parser`, so we do not expect any disruptions
from these changes.

A detailed summary of the changes are listed below.

- `assert()` is now used to detect same files with different IO units.

  Detection of reopenend files of the same name but different IO unit
  has been changed from `MOM_error(FATAL, ...)` to `assert()`, to
  reflect that this should be a logically impossible result.

- Bugfix: Reopened files are now reported to all PEs.

  If an open file is re-opened, then only the root PE will detect this
  and will `return` immediately.  However, the others will proceed into
  `populate_param_data` and will get stuck in a broadcast waiting for
  root.

  We fix this by communicating the reopened state to all PEs and allow
  all ranks to return before re-processing the data.

  Note that this could also be resolved by allowing all ranks to track
  IO unit numbers, but for now we do not attempt to change this
  behavior.

- `newunit=` used to generate parameter file IO unit

  The parameter IO unit is now generated by `newunit=` rather than an
  explicit search for an unused IO unit.  Note that this is a Fortran
  2008 feature.

  Testing around available IO units has also been removed.

- Removal of generic IO error handling

  Generic "IO error" tests, and corresponding `err=` arguments, have
  been removed in most cases.  We now rely on the Fortran runtime to
  provide diagnostics on these errors, which should typically exceed any
  information that MOM6 could provide.

- Removal of purported `namelist` support

  There were several blocks of code provided to support namelist syntax,
  but did not appear to be working, nor was there any known instance of
  it being used by anyone, so it has been removed.

- `#define/undef/=` syntax testing across ranks

  Previously, only the root PE would test for consistency of the
  #define-like syntax, even though all ranks have this information.
  This required a second, awkwardly placed syntax test later in the
  subroutine.

  This test is redefined to run over all ranks, and the subsequent test
  has been removed.

- `define/override` test reordering

  The `found_override` test when coupled to a `#define`-like declaration
  was unreachable due to the presence of an even stronger test related
  to valid syntax.

  This test has been moved to provide more detailed information about
  the nature of the error.

- `link_parameter`, `parameter_block` defined as private

  Internal derived types of `MOM_file_parser` are redefined as private.
  This preserves the integrity of instances of these types, and also
  prevents creation of implicit object code required to access them
  externally.

- Removal of `static_value` from `get_param` interface

  The `static_value` argument of `get_param` has been removed, since it
  is functionally equivalent to `default`.  While this is an API change,
  there is no known case of anyone using this argument.

- The `param_type%doc` fields are now properly deallocated after closed.

- Quotes have been added around some filename error warnings, to help
  detect issues related to whitespace.

- `any_across_PEs` and `all_across_PEs`

   New functions for calling `any()` and `all()` across PE ranks have
   been added.  Behavior is in line with other functions, such as
   `min_across_PEs`.

(+) Refactor of MOM_file_parser

  Added a number of options to MOM_ALE to mimic options that are found in
Hycom.  By default, all answers are bitwise identical, but there are several
new runtime parameters.  The code changes include:

 . Added the ability to use a different remapping scheme for velocities than
   for tracers, using the new runtime parameter VELOCITY_REMAPPING_SCHEME

 . Added the new runtime option PARTIAL_CELL_VELOCITY_REMAP to use partial cell
   thicknesses for remapping at velocity points, which triggers a call to the
   new internal routine apply_partial_cell_mask.

 . Added the new internal routine mask_near_bottom_vel to allow MOM6 to mimic
   Hycom in zeroing out the velocities in thin layers in a bottom boundary
   layer with a thickness given by the new runtime parameter
   REMAP_VEL_MASK_BBL_THICK, while the definition of thin is specified by
   REMAP_VEL_MASK_H_THIN.  Setting these to be negative (as is the default)
   avoids this.

 . Modified the interface to remap_all_state_vars to take just the ALE_CS, which
   then provides the remapping control structure that is appropriate for the
   tracers or velocities, rather than also passing this in as an argument.

 . Eliminated some unnecessary internal variables, and added others to be
   more explicit avoid array syntax copies in arguments.

  Enforce a minimum thickness of 0.5*Angstrom in the mixed_layer_restrat
routines.  The streamfunctions in these routines attempt to limit the
thicknesses to exceed Angstrom, but they can be less than this due to roundoff.
The new limit prevents thicknesses from becoming negative when Angstrom is set
to 0, but should not change any answers for test cases with positive values of
Angstrom.  Also added some comments describing variables and their units and
simplified the OMP directives.

  Also corrected error messages in MOM_diabatic_aux.F90 to identify the file or
module where these messages come from, and modified an error message in
applyTracerBoundaryFluxesInOut so that it is written if the localized fault does
not happen to occur on the root PE.

  All answers in the existing MOM6-examples regression suite are bitwise
identical.

This patch redefines `h_new` to match the shape of a center-point rather
than a v-face point.

Dynamic memory builds would have been unaffected by this, and older GCC
(~7.3.0) compilers appear to have not objected to the shape mismatch,
but this raised an error in newer GCCs (~9.3.0).

Since this redefines `h_new` from zero-based to 1-based indexing in the
y-axis, answer changes are very possible.

bugfix: static h_new shape remaining_transport_sum

  Stop logging the deprecated run-time parameter NEW_SPONGES, and always log
INTERPOLATE_SPONGE_TIME_SPACE as if NEW_SPONGES were not used.  This commit will
address MOM6 issue #46, which can be closed it is accepted.  This will change
the MOM_parameter_doc entries in some cases, but all answers are bitwise
identical.

- Enabling pipelines for SIS2 required making the MOM6_SRC macro mandatory.
  It was set in .gitlab-ci.yml for all targets except nolibs build.

+(*)Fix dimensional rescaling with HOMOGENIZE_FORCINGS

  This commit adds the ability to specify three remapping options derived from
Hycom's hybgen code.

 - Adds the new module MOM_hybgen_remap, which contains the new subroutines
   hybgen_plm_coefs, hybgen_ppm_coefs, and hybgen_weno_coefs

 - Adds code to handle PLM_HYBGEN, PPM_HYBGEN and WENO_HYBGEN as valid entries
   for specifying the ALE remapping schemes.  Also added descriptions of units
   to some internal remapping variables.

 - Adds the optional argument PCM_cell to regridding_main, remapping_core_h, and
   remap_all_state_vars to specify layers that should use piecewise constant
   remapping, regardless of the overall remapping scheme, to follow the approach
   used in Hycom.  ALE_main uses these new optional arguments, although until
   the Hybgen regridding code is added, they will always be set to false.

 - Makes 7 character strings longer in 5 files to accommodate the new
   remapping options.

All answers are bitwise identical, but there are changes to some entries in the
MOM_parameter_doc files.

+Add Hybgen remapping options

  Added the capability to do Laplacian diffusion of interface heights without
regard to the density of layers.  This is accomplished by the new internal
subroutine add_interface_KH, and is controlled by the new runtime parameters
KH_ETA_CONST and KH_ETA_VEL_SCALE.  The new interface diffusivities are subject
to CFL bounding when combined with the previous isopycnal slope diffusivities,
with the two diffusivities reduced proportionately when the CFL limits are
exceeded.  Also simplified many openMP directives in MOM_thickness_diffuse.F90,
and fixed a number of spelling errors or other problems in comments, including
adding units to some variables.  By default, all answers are bitwise identical,
although there are new runtime parameters a new internal subroutine.

  This change avoids an unnecessary copy of the tracer registry in
advect_tracer().  This should be more efficient, and it should facilitate the
development of improved tracer budget diagnostic capabilities.  It also adds
comments describing a few internal variables and their units.  All answers and
output are bitwise identical.

  Corrected the dimensional rescaling in RGC_initialize_sponges().  Also added
comments documenting the units of all the real values in this module.  Only
dimensional rescaling factors are changed, and the code should now reproduce
for different values of dimensional rescaling factors.  All answers in the
MOM6-examples regression suite are bitwise identical.

  Corrected the license on RGC_initialization.F90 to refer to the same
LICENSE.md file as the rest of the MOM6 code, with the explicit written consent
of the contributors to this file.  Also changed the local variable "mod" into
"mdl" in this module to avoid some of the problems we had previously found with
some compilers from reusing an intrinsic function name for a variable name.  All
answers are bitwise identical.

Update main to dev/gfdl

  Fixed the bug noted in issue #72 and excessive loop sizes noted in the
unresolved comments in a recent commit, and cleaned up other aspects of
MOM_hor_visc, mostly related to the code that is exercised when USE_GME=True.

 - Fixed the bug with the wrong arrays being used when ADD_LES_VISCOSITY=True
   that was noted in MOM6 issue #72.

 - Corrected some of the overly large loop extents that were noted in unresolved
   comments about MOM6 PR #65.

 - Only log USE_KH_BG_2D if a Laplacian viscosity is used.

 - Use extra calculations in the halos and marching in to avoid unnecessary halo
   updates in smooth_GME if there multiple smoothing passes.

 - Corrected the capitalization of some horizontal indices to follow the MOM6
   convention for indicating the horizontal staggering position.

 - Collected the post_data calls for diagnostics with use_GME with the other
   post data calls to collocate some of the potential inter-PE synchronization
   points.

 - Fixed the indentation of some expressions that were using less than 4 points
   for some continuation lines.

 - Eliminated several unused variables, and fused some loops to allow 2-d
   variables to be replaced with scalars.

  With this PR, answers can change when ADD_LES_VISCOSITY=True and there is a
Smagorinsky or Leith Laplacian viscosity and there is a nonzero background.
One run-time parameter is no longer logged if LAPLACIAN=false, so there are
changes to the MOM_parameter_doc files.  All answers in the MOM6-examples test
suite are bitwise identical.

  Minor code cleanup in response to the code review from Gustavo Marques.  In
particular, this introduces a roundoff-level answer changing code simplification
for code that is only exercised if USE_GME=True.  In all other cases the answers
are bitwise identical.

  Added optional arguments to framework code to permit the restart files to work
with dimensionally unscaled variables.  All answers are bitwise identical, but
there are new optional arguments for multiple public interfaces.

 - Added a new conversion factor element, conv, to the field_restart type to
   specify how a field is to be unscaled before writing, and whose reciprocal is
   used to scale restart fields as they are read in.

 - Added the new optional argument conversion to each of the
   register_restart_field() and register_restart_pair() routines, which is used
   to specify how each field will be rescaled when it is written to or read from
   a restart file.  The default is the same as setting this to 1.

 - Added the new optional argument unscaled to fix_restart_scaling() and
   fix_restart_unit_scaling() to reset the unit conversion factors that will be
   save to the restart files to 1, consistent with writing out fields without
   scaling to the restart files.

  Revised the MOM6 code so that the output restart files are always exactly the
same as they would be if no dimensional rescaling is applied.  The input restart
files can still have rescaling, so files written by previous versions of the
code still work exactly as before.  This does change the output, in the sense
that the restart files are unscaled and some units documents in the restart
files are corrected, but the solutions themselves are bitwise identical.  Also,
there are new (non-optional) unit scaling type arguments to several routines.
The specific changes in this commit include:

 - Added conversion factors to all register_restart_field() or
   register_restart_pair() call for variables that are subject to dimensional
   rescaling.

 - Revised the calculation of restart rescaling factors to reflect the rescaling
   that now occurs directly in the restore_state() call.

 - Added new US arguments to register_restarts_dyn_split_RK2(),
   MEKE_alloc_register_restart(), set_visc_register_restarts()

 - Added a new GV argument to mixedlayer_restrat_register_restarts()

 - Used the new unscaled argument in calls to fix_restart_scaling() and
   fix_restart_unit_scaling()

 - Revised several calls to register_restart_field() to avoid using the variant
   that uses a var_desc type, eliminating the need for some modules to reference
   the vardesc type or the routine var_desc().

 - Added or corrected unit descriptions to the comments describing a few
   variables.

 - Noted a few probably bugs in comments with ###, but did not fix them.

+Stop logging NEW_SPONGES

* In MOM_ice_shelf_dynamics.F90 changes are  made to calc_shelf_visc() and calc_shelf_driving_stress() to account for an irregular quadrilateral grid. In MOM_ice_shelf_initialize.F90 changes are made to initialize_ice_thickness_from_file() to correct masks at initialization.

* Changed indentation

* Changes are made to 'calc_shelf_visc()` to make computations of the velocity derivatives rotation-invariant. Changes in `update_ice_shelf()` utilize G%IareaT(:,:) instead of 1/G%areaT(:,:).

* Allocate shelf_sfc_mass_flux for dynamic ice shelf configurations

* Conditional allocation of ice shelf fluxes and data override call for sfc mass flux

* get rid of excessive line length

* call data override init for ice shelves

* A new subroutine change_thickness_using_precip() changes the ice thickness in response to the atmospheric mass flux

* Conditional calls to data_override

  - Introduces a new flag: DATA_OVERRIDE_SHELF_FLUXES, which
    if set to True, and ACTIVE_SHELF_DYNAMICS is also True, will
    enable data_override capability for the surface mass deposition
    (only avaialble for MOSAIC grid types)

* Changes are made to extrapolate_metric() in MOM_grid_initialize.F90 to avoid negative values at the southern boundary of the domain. This is done for grids extending over Antarctica.

* Reversed changes in MOM_grid_initialize.F90 to debug a regression test.

* Uncommented lines reading the ice velocity from file

* Changes to scaling factors and conversions of the 'sfc_mass_flux' parameter

* Remove unnecessary diagnostics for ice shelves from MOM_forcing_type

* Correct rescaling for data override of ice shelf sfc mass fluxes

* Add diagnostic rescaling for ice shelf mass fluxes

* Uncommented assignement of 'area_shelf_h(i,j)=G%areaT(i,j)' in MOM_ice_shelf_initialize.F90 and modified the four halo updates in MOM_ice_shelf.F90.

Co-authored-by: Matthew Harrison <[email protected]>

* reworked algebra corner spread, rotation parenthesis

* replace .lt. by <

Co-authored-by: Raphael Dussin <[email protected]>

  Added the ability to detect and handle multiline parameter specifications that
are longer than INPUT_STR_LENGTH after the lines have been concatenated.  Many
instances where character strings are set to length INPUT_STR_LENGTH are
replaced by param_file_type%max_line_len.  This commit partially addresses MOM6
issue #75.  All answers and output are bitwise identical in cases that worked
previously, but there is a new element in the publicly visible param_file_type,
and some cases where there had previously been errors due to very long input
specification may now work.

  Make the lines of the param_file_type individually allocatable to support
arbitrarily long input lines.  This required the introduction of a new private
type to hold the actual allocatable array, so the lines are now an array of this
type.  All answers are bitwise identical.

  Use a character buffer to minimize the memory footprint associated with
reading the input parameters on root_PE and broadcasting it to all PEs.  All
answers are bitwise identical.

  Now that INPUT_STR_LENGTH is only used as the size of a single temporary
character string, the maximum permitted line length can be increased to a larger
value without unduly impacting the memory footprint of the model, so it has been
increased to 1024.  If this is not large enough, considering that MOM6 also
supports line continuation, there is a bigger problem than this limit.  Some
message strings were also changed to be allocatable, while others have sizes
that are set using param_file_type%max_line_len.  Also eliminated the hard-coded
internal parameter all_PEs_read, which had a note that it should have been
eliminated in about 2010 (before MOM6 existed).  All answers and output are
bitwise identical.

This patch introduces some limited support for error handling in a POSIX
environment.  A partial interface to the POSIX system calls is included
as part of the implementation of this feature.

An internal global logical flag, `ignore_fatal` is added to the
`MOM_error_handler` module which controls the behavior of `FATAL`
errors.  If true, then `FATAL` errors trigger a signal (here `SIGUSR1`)
which call a handler which restores the program to the earlier state.

This change allows us to test `FATAL` errors at runtime without stopping
program execution, so that we can cycle through multiple errors.

This feature is toggled with two new functions, `disable_fatal_errors()`
and `enable_fatal_errors()`.

This is implemented in part using POSIX system calls from the C standard
library (libc), using the C interoperability layer.  Details which are
unspecified by the POSIX standard, such as the definitions of signals or
the datatype size, are defined in a header (`posix.h`) which can be
optionally configured at compile-time.

The header defaults to use typical Linux POSIX values, and may be
configured by autoconf for more exotic platforms in a later patch.

The previous state is set and restored by the `setjmp`/`longjmp` (or
`sigsetjmp`/`siglongjmp`) functions.  Although these functions behave as
expected on all tested platforms, a code which uses these functions is
not standards-conforming.  If issues arise in the future, error handling
may not be available on those platforms, and compile-time configuration
may be required to disable these features.

No existing code is using these features, but are expected to be part of
a future patch.

Error handler API:
* `disable_fatal_errors`
* `enable_fatal_errors`

New POSIX API:
* `chmod`
* `signal`
* `kill`
* `getpid`
* `getppid`
* `sleep`
* `setjmp`
* `sigsetjmp`
* `longjmp`
* `siglongjmp`

New derived types:
* `jmp_buf`
* `sigjmp_buf`

New interface (for signal handlers):
* `handler_interface`

New compile-time macros:
* `JMP_BUF_SIZE`
* `SIGJMP_BUF_SIZE`
* `SIGSETJMP_NAME`
* `POSIX_SIGUSR1`

MOM6: +(*)Write unscaled restart files.

Co-authored-by: Raphael Dussin <[email protected]>
Co-authored-by: Marshall Ward <[email protected]>

  Rearrange how the vertical_coordinate file is written, to avoid an unnecessary
repetition of the writing of this file, and to prepare for the Hybgen grid
generator code to write its own version of this file.  The specific changes
include:

 - Added the new subroutine write_regrid_file() to write the vertical_coordinate
   file when in ALE mode.

 - Relocated the code to write the vertical_coordinate file in ALE mode
   from ALE_writeCoordinateFile to write_regrid_file()

 - Moved the call to write_vertgrid_file() into initialize_MOM(), and do not
   call it when USE_REGRIDDING=True, since in this case this file will be
   rewritten again when write_regrid_file() is called.

 - Eliminated the write_geom and output_dir arguments to MOM_initialize_coord()

All answers and output are bitwise identical.

  Add the optional conversion argument to register_diag_scalar, which works
analogously to the argument that is already available in register_diag_field.
All answers are bitwise identical, but there is a new optional argument in a
public type.

  Use conversion argument during registration and tmp_scale argument in the
calls to global_area_mean for averaged fields in the MOM_forcing_type and
MOM_diagnostics modules.  Also added or corrected some descriptive comments and
fixed an extra unit conversion factor in one recently added calculation of ustar
that was likely not used.  All answers are bitwise identical.

  Use the tmp_scale argument in calls to global_area_integral to simplify some
global integrals, and added dimensional rescaling for the time_step element of
the ice_shelf control structure.  Also corrected some dimensional descriptions
in comments.  All answers are bitwise identical.

  Rescaled the new_depth variable in apply_topography_edits_from_file, and used
source arguments to initialize 6 allocated variables.  All answers are bitwise
identical.

  Replaced a call to global_area_mean with a call to global_area_integral in a
diagnostic calculation in the internal_tides module.  Also added descriptions of
the dimensions of a number of variables in this module, or eliminated the
dimension description of several integers (for which units make no sense).  Only
an unused debugging diagnostic is changed, and all solutions are bitwise
identical.

  Eliminated units arguments for logical, integer, or character string get_param
calls where the units make no sense.  In some other cases, calls were slightly
revised to place the units and scale arguments on the same line for easier
detection of inconsistent settings.  All answers are bitwise identical, but
the MOM_parameter_doc.all files for some test cases are corrected.

  Revised get_param calls to put the units and scale arguments on the same line,
to help detect inconsistent settings.  All answers and output are bitwise
identical.

  Eliminated a commented-out get_param call and use a source argument to
initialize some arrays.  All answers and output are bitwise identical.

  Corrected the units in some get_param calls in the dumbbell test case.  All
answers are bitwise identical, but there will be minor changes in the
MOM_parameter_doc files for the dumbbell test case.

Co-authored-by: Robert Hallberg <[email protected]>