Ensure compatibility with pandas CoW

[glatterf42]

Closes #804.
This PR temporarily enables the pandas Copy-on-Write behaviour that will become standard with pandas 3.0. With this, I ran all tests locally to study the warnings we would get from this: only two files were affected, add_year/__init__ and macro. Then, I fixed all of these warnings as suggested in the migration guide, so we might be good with this repo (and all underlying ixmp functions called from the test suite here) :)

How to review

• Read the diff and note that the CI checks all pass.
• Study CI output to see if any pandas Warnings wrt CoW came up that were skipped locally.
• Fix new Warnings (if any).
• Remove the TEMPORARY commit to avoid enabling pandas CoW for all users already.

PR checklist

• Continuous integration checks all ✅
• [ ] Add or expand tests; coverage checks both ✅
• [ ] Add, expand, or update documentation.
• Update release notes.

[codecov]

Codecov Report

Attention: Patch coverage is 71.59091% with 25 lines in your changes missing coverage. Please review.

Project coverage is 95.6%. Comparing base (0cc8102) to head (4913c24).
Report is 55 commits behind head on main.

Files	Patch %	Lines
message_ix/tools/add_year/__init__.py	70.9%	25 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##            main    #842     +/-   ##
=======================================
+ Coverage   95.4%   95.6%   +0.1%     
=======================================
  Files         46      46             
  Lines       4354    4339     -15     
=======================================
- Hits        4156    4149      -7     
+ Misses       198     190      -8     

Files	Coverage Δ
message_ix/macro.py	96.7% <100.0%> (+<0.1%)	⬆️
message_ix/tools/add_year/__init__.py	66.8% <70.9%> (+0.6%)	⬆️

[khaeru]

So long as CI passes, we are good to go.

I made some inline comments that maybe would be irrelevant if #494 will refactor the affected code, so they are just FYI.

[khaeru]

#494 was intended to pick up and improve many of these usages. Here, I suspect .xs() could be used. No need to actually make that change in this PR, just noting for later reference.

[glatterf42]

Thanks for the note :)

[glatterf42]

Not sure why codecov/patch registers a diff hit, we test the same functions as before. So I'm fine with the hit if that comes from additional lines that reduce complexity. I'll rebase to exclude the TEMPORARY commit and merge once the tests pass (again).