TST: start testing on Python 3.11

[neutrinoceros]

PR Summary

Fix #3954
This is expected to fail until upstream is sufficiently stable. Currently the limiting factor is kiwisolver, but it looks like it'll be fixed in the coming days.

[neutrinoceros]

A new release of kiwi solver is expected to drop next week (see nucleic/kiwi#142)
I'll rebase and fix the conflict then.

[neutrinoceros]

Ok here's an actual incompatibility on our side:
yt currently fails to build on CPython 3.11.0b3 with the following error

/opt/hostedtoolcache/Python/3.11.0-beta.3/x64/include/python3.11/pyport.h:61:20: error: ‘nullptr’ was not declared in this scope

This happens because the cykdtree c extension is explicitly requesting -std=c++03, but python/cpython#92253 makes C++11 de facto required (nullptr wasn't a builtin before that)

I'm not sure wether the bumped requirement is intended on the CPython side or not.

[matthewturk]

So could we bump to c++11 in our build, or remove the line entirely when on python 3.11?

[neutrinoceros]

I think the reason why certain extensions require C++03 is that we wanted to support old compilers like GCC 5 (see #2939)
Ideally we should make this conditional for Python <= 3.10, but I'm don't know how.
Otherwise we may just embrace C++11 as the minimal requirement. It's probably more reasonable now than two years ago anyway, especially since our more rapid release cadence makes it less relevant to install from source.

[matthewturk]

I have some uncertainty about doing this without checking some supercomputer installations, which are notorious for having out of date compilers. But, we use c++11 for the bitmap index stuff, right?

[neutrinoceros]

Actually we clarified that making C++11 required wasn't intended, and there's a fix in the making python/cpython#93784
So we should be able to compile without any change on our side when this ships !

[neutrinoceros]

After a long delay, Python 3.11.0b4 is available, let's try this again

[neutrinoceros]

not quite. actions/python-versions#177 is still pending

[neutrinoceros]

It compiles !
Now we're seeing 4 occurrences of the same problem, for instance:

DeprecationWarning: It is deprecated to return a value!=None from a test case (<bound method SimpleVRTest.test_simple_vr of <yt.visualization.volume_rendering.tests.test_simple_vr.SimpleVRTest testMethod=test_simple_vr>>)

This seems to be a pytest deprecation. I don't understand why it's only showing up now but a quick inspection seems to indicate that in all 4 cases, the return value of these tests are actually not used, so it shouldn't hurt to just remove them.

[neutrinoceros]

This is now ready !

@matthewturk can you take a look ? The sooner this gets in the sooner we'll be able to report any problem with 3.11

[matthewturk]

Hmm, I thought for a minute that removing the return would make this harder to debug, but now I'm not sure that it matters. So I'm OK with this.