fix RaisesGroup(strict=...) in tests, removed in trio 0.27

[mikenerone]

Trivial fix replacing RaisesGroup(strict=False) with RaisesGroup(flatten_subgroups=True) in tests (strict was deprecated in Trio 0.26 and removed in Trio 0.27 in October 2024).

Also fixes a couple of tiny docs-related issues that were breaking the docs build due to recent external changes, and thus failing CI:

1. RTD now requires an explicit configuration setting specifying the config file.
2. Sphinx no longer requires an explicit call to get_html_theme_path(), and in fact warns WARNING: Calling get_html_theme_path is deprecated. If you are calling it to define html_theme_path, you are safe to remove that code. (and CI treats warns as errors).

Fixes #154

[mikenerone]

I'm failing to grok why CI is failing for Python 3.12 under MacOS. I've run the tests that way locally and can't reproduce. @jakkdl or anyone else, any ideas?

[jakkdl]

I have no clue, @graingert do you have any ideas?

[graingert]

I'm afk for a while, I usually debug this sort of thing using objgraph. You might need to use https://github.com/marketplace/actions/debugging-with-tmate if you can't repeat locally

[noahbkim]

Thinking this is spurious as I cannot reproduce locally on my MacBook with Python 3.12, have you tried rerunning the checks?

[jakkdl]

Thinking this is spurious as I cannot reproduce locally on my MacBook with Python 3.12, have you tried rerunning the checks?

yeah 😕

[jakkdl]

but it's unrelated both to this and #156, so I think we can mark it xfail and try to address it in a separate PR

[mikenerone]

...I think we can mark it xfail and try to address it in a separate PR

If I just xfail it, the test will fail when running locally (i.e. because it succeeds where we're expecting a failure). Maybe I should rather skip it for Python 3.12/MacOS.

More fundamentally, I'm concerned that something really might have changed with recent Python 3.12, and the failure is somehow legitimate, and we might end up with a broken release next time if we forget to get to the bottom of this before then.

[mikenerone]

Well, nevermind, whatever it was resolved itself. I've rerun the tests (again) twice today, and it passed both times. I would guess that "they" fixed something in the Python3.12/MacOS test environment.