Fix: apply_ufunc with exclude_dims and vectorize

[mathause]

• Closes error for apply_ufunc with exclude_dims and vectorize #3890
• Tests added
• Passes isort -rc . && black . && mypy . && flake8
• Fully documented, including whats-new.rst for all changes and api.rst for new API

This would be pretty useful when using apply_ufunc for e.g. statistical tests that require vectorize=True and can consume data of unequal length (i.e. uses exclude_dims). This should also extend to dask arrays with dask="parallelized" once #4060 is implemented.

[shoyer]

Thanks for working on this! What do you think of my suggestion of putting this in the to_gufunc_string method rather than in the data model for _UFuncSignature? That seems a little less invasive to me.

[mathause]

Sounds good. I can move _enumerate into to_gufunc_string. I think it makes sense to wait for #4060.

[mathause]

Inplemented @shoyer's suggestion to add exclude_dims as optional argument to to_gufunc_string. The readthedocs error is a timeout so should be fine to review.

[keewis]

The readthedocs error is a timeout

Unfortunately not. The build timed out after the error and the traceback is visible. My own setup doesn't time out on errors, so I asked support to help fix that. They removed stale memory constraints (which is why the traceback is visible), but somehow the build still times out on errors.

In this case pandas has a minimum version requirement for xarray, but the detected version is 0.1. I'm not sure why version detection works for me but not for you, maybe because you didn't update the tags on your repo in a long time?

Edit: to re-trigger RTD, we have to close and reopen the PR; they don't allow rerunning a PR build

[mathause]

Thanks keewis! I don't think I would have figured that out. I pushed the tags to my fork.

[mathause]

But unfortunately this doesn't help :( If you have another idea I am all ears. But anyways it should not stop anyone from reviewing.

[keewis]

on my own setup the build completes: https://readthedocs.org/projects/xarray-keewis/builds/11696307/ So this is definitely an issue with the setup, but I don't know how to fix that...

[kmuehlbauer]

@keewis In the xarray setup there is --prune-tags. In your setup this is not there. Not sure, if this might be connected.

[keewis]

thanks, @kmuehlbauer, good spot. However, that seems to come from the PR build (see also https://readthedocs.org/projects/xray/builds/11695402/, which doesn't fail), so I guess there has to be something in the code that changes the version.

[dcherian]

Very clean. Great work!

[keewis]

the merge seems to have fixed it. Not sure why it didn't work before.

[mathause]

Thanks @keewis!