Replace resolved numpy workaround with dask workaround

[glatterf42]

This PR tries to resolve the CI test failures by not using version 2024.11.0 of dask, which is causing issues with genno.

The most preferable solution would be to adapt genno to this dask version, but it is unclear how much effort this would be and how long this would take. So for now, I'm simply imposing that we don't use that particular version (note: 2024.11.1 would be allowed, hoping that this might resolve the incompatibilities).
If this workaround is successful and we agree we want to employ it for now, we should also adopt it in the nightly workflow.

How to review

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

PR checklist

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

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 95.5%. Comparing base (9d9f5bd) to head (d177d2a).
Report is 1 commits behind head on main.

Additional details and impacted files

@@          Coverage Diff          @@
##            main    #887   +/-   ##
=====================================
  Coverage   95.5%   95.5%           
=====================================
  Files         46      46           
  Lines       4340    4340           
=====================================
  Hits        4148    4148           
  Misses       192     192           

[khaeru]

Thanks!

I would only suggest to use dask < 2024.11.0 instead of dask != 2024.11.0.

The reason is that, until or unless we make a report to dask upstream and they choose to fix it, we should expect that future releases of dask (e.g. 2024.11.1, 2024.12.0, 2025.1.0) will also cause the same errors.

Using < avoids any need to adjust the workaround.

[glatterf42]

Thanks, sounds good :)
I can open a PR similar to this one on message-ix-models, too.