Multi cluster support

Added option for multi-cluster support.
Thanks @Lafond-LapalmeJ for the PR!

Fix state filtering of sub jobs

Fixes error where state filtering would remove information about sub-jobs if the states didn't match.

Patch changes into last python 3.6 supported version

Dropped in all tests and changes. Had to remove custom filenames due to regex changes.

Should prevent issues due to old bugs on systems with old python versions.

Fix memory with multiple tasks

Fixed estimation of memory efficiency with multiple tasks.
A few source code improvements.

Thanks @biermanr for help coding and @angel-devicente for providing a test case.

Handle Pipes in jobname

Changed sacct delimiter to ^|^ so job names can contain a pipe without breaking.
Using reportseff's parsable option will still output a single pipe.

Change color defaults

Change default to let click decide how to pipe color.

Added a discussion to the readme of how to handle the changes.

Thanks @braffes for opening the issue at the PR!

Hotfix, remove print statement

v2.7.3

fix: remove print statement

More parsing edge cases

Added more help for using the slurm-format option.
Handled cases where total time from the admin comment could be 0.

Thanks @braffes for the bug report.

Slurm format option

Added an option to specify the slurm output file format, accepting the format used by sbatch. This addresses using output names like %j.out which previously were ignored.

• %j or %A can be used to find the job id
• For array jobs, %A_%a must be present
• One of the above needs to be present or a value error is thrown as the jobid cannot be determined.
• Other, sequential format tokens are then collapsed into a single .*, e.g. %n%x_%j.out -> .*_<JOBID>\.out
• Any regex tokens in the format are escaped prior to processing and compiling.

Handle more missing GPU information

Added more guard clauses for calculating GPU information from AdminComment.