Support Sphinx 5

[jarrodmillman]

I am still in favor of dropping the upper bound (see #661). But while we think about how to handle that PR, it seemed like supporting the two most recent major Sphinx releases was less controversial. Sphinx 5 is scheduled for release in a couple weeks.

Assuming this is acceptable, I would like to request a pre-release of pydata-sphinx-theme soon. That would help us test numpydoc on sphinx 5. Then, ideally, I would like to coordinate new releases of numpydoc and pydata-sphinx-theme soon after Sphinx 5 is released.

I am more than happy to help with the release. If that would help, I suspect I would need access to this repo and PyPI.

[choldgraf]

+1 from me, provided that:

• Our test suite is passing for the latest major version of Sphinx (now I'm wondering if we need to test both Sphinx 4 and Sphinx 5 (or 3, 4, and 5) in our test suite? That sounds like a combinatorial tests explosion though...)
• Double check that none of the demo documentation pages have any obvious problems in them when the docs are built with Sphinx 5

Also +1 on a pre-release quickly, @jarrodmillman I am happy to provide you maintainer rights on the repository and maybe we can use this as an opportunity to improve the release docs if you hit any snags.

[jarrodmillman]

The test suite already tests on 5.0.0.b1:

There is an error (that pip currently ignores) since myst-parser also upper bounds Sphinx.

I will look into the demo documentation pages.

[jarrodmillman]

I just built the docs on my local machine using Sphinx 5.0.0.b1 and compared the output visually. The only difference I saw was in the docstring for asyncio.gather(*coros_or_futures, loop=None, return_exceptions=False), which was the result of building the docs with Python 3.7 vs. 3.9. So as far as I can tell from manually comparing the output from several pages of the demo site, pydata-sphinx-theme works fine on Sphinx 5.0.0.b1.