Miscellaneous packaging improvements

[khaeru]

• Use pyproject.toml for packaging, instead of setup.{cfg,py}.
• Replace usage of deprecated pkg_resources with importlib.metadata/importlib_metadata (Python 3.7 only).
• Various housekeeping and clean-up of the CI workflows.
• Use pytest-xdist to parallelize tests (n=2) on GHA runners; this shortens run time of individual jobs by up to 5 minutes.
• Cherry-pick two commits (by @glatterf42 originally at Test & improve perfomance in .tools.add_year #494) adjusting for pandas 2.0.0.

How to review

Read the diff and note that the CI checks all pass.

PR checklist

• Merge Miscellaneous updates for 2023-W15 ixmp#476, required to address an issue with the codecov dependency.
• Continuous integration checks all ✅
• Add or expand tests; coverage checks both ✅
• Add, expand, or update documentation. N/A
• Update release notes.

[codecov]

Codecov Report

Merging #703 (d2c461b) into main (03a0eb5) will decrease coverage by 0.5%.
The diff coverage is 100.0%.

@@           Coverage Diff           @@
##            main    #703     +/-   ##
=======================================
- Coverage   94.9%   94.4%   -0.5%     
=======================================
  Files         42      43      +1     
  Lines       3366    3432     +66     
=======================================
+ Hits        3195    3241     +46     
- Misses       171     191     +20     

Impacted Files	Coverage Δ
message_ix/__init__.py	100.0% <100.0%> (ø)
message_ix/core.py	95.1% <100.0%> (ø)
message_ix/testing/__init__.py	99.6% <100.0%> (ø)
message_ix/tests/test_macro.py	99.4% <100.0%> (ø)
message_ix/tests/test_tutorials.py	100.0% <100.0%> (ø)
message_ix/tools/add_year/__init__.py	66.2% <100.0%> (ø)

... and 79 files with indirect coverage changes

[khaeru]

@glatterf42 #683 led to conflicts in test_tutorials.py; please wait for me to rebase & deconflict before reviewing.

[glatterf42]

Sorry, I didn't realise it would. I'll review once you are done.

[glatterf42]

Looks good to me, only one question and a few very minor remarks from my side.

[khaeru]

NB the codecov/patch check failure occurs because a few lines are changed in .tools.add_year, but that code remains untested. The tests will be added by #494, from which those changes were cherry-picked.

[glatterf42]

Do we now have to remove those commits from #494 or does git recognize that we included them here already once we merge both?

[khaeru]

Do we now have to remove those commits from #494 or does git recognize that we included them here already once we merge both?

Yes, they can be removed in a few steps:

1. Check out that branch.
2. git rebase -i HEAD~10 —replacing "10" with some number that goes back far enough to include those commits.
3. Edit the TODO list in the editor that is opened. Put "d" or "drop" on the start of the line for those commits that have been cherry-picked; leave the rest the same.
4. git rebase main; this will move the branch to the head of main, which already contains the cherry-picked commits.

Another way is to just do (1) and (4). During the rebase, git will complain, twice, that the cherry-picked commit "is (now) empty", because all changes they contained are already on main. It will prompt you to do git rebase --skip each time.

[glatterf42]

This is mostly just FYI: I've applied your suggestion to #494 and ran (1) and (4) after updating my local main. The process was slightly different for me, then: git immediately told me that it skipped the two previously applied commits. It still found two merge conflicts, however, which I had to resolve manually (both about the beginning of the definition of the interpolate_1d function). After that, I had to force push the branch because my initial updates to main were not yet included in the upstream branch.

[khaeru]

In this PR I forgot to adjust this line in the "nightly" CI workflow:

message_ix/.github/workflows/nightly.yaml

Line 47 in a824be9

cache-dependency-path: "**/setup.cfg"

…to refer to pyproject.toml, mirroring the change made to the "pytest" workflow. As a result the workflow fails, e.g. https://github.com/iiasa/message_ix/actions/runs/4728750039