update python versions and dependencies in pyproject.toml

[Jhsmit]

fixes #19

is it correct that ultraplot is python version >= 3.9?

if so, the previous 'importlib-metadata; python_version>"3.8"', can be removed since python 3.8 is no longer supported.
also updated classifier and requires-python

[cvanelteren]

The build logs indicate that mpl 3.9 is not compatible with 3.8. It states that the ABI for python 3.8 is not compatible with mpl 3.9. To be precise this is the reason raised:

 ├─ matplotlib >=3.9,<3.10 * is installable with the potential options
      │  ├─ matplotlib [3.9.1|3.9.2] would require
      │  │  └─ python_abi ==3.9 *_pypy39_pp73, which requires
      │  │     └─ python =3.9 *_73_pypy, which can be installed;

This would mean that UltraPlot also drop support for python 3.8 -- although in practice it would probable work fine.

[cvanelteren]

Made a wiki entry proposing support process moving forward. Open to hearing feedback from the community on this; whether adopting an N-2 feedback model would work for the users etc.

[cvanelteren]

I think it would be best to release a minor version update (v1.0.1) and from the next major version (v1.1) drop support for python <3.11.

[Jhsmit]

You would drop py 3.10 already?
I personally dont see reasons to aggressively drop python version support. Yes you get to use newer python versions sooner but for a project like ultraplot I would try to support an as large an audience as possible.

Arguments for keeping py 3.10:

• it isn't end of life yet: https://devguide.python.org/versions/
• matplotlib still supports python 3.10
• I personally still use python 3.10 in several projects and would like to use ultraplot there / I'm considering adding it as dependency.

[cvanelteren]

You are absolutely right. I have no qualms keeping it in until it's end of life.

[Jhsmit]

Ok,
Your approach to matrix building for python versions is interesting, I'm wondering why you use the script to get versions instead of write them in the workflow yaml directly? single source of truth and all that?

In any case, it seems to work, PR is fine to merge from my side

[cvanelteren]

This way it directly grabs the supported version from toml and we can set it an forget it 😬

[cvanelteren]

Don't worry about the red. I have added the parameter to force similarity checks between plots

There are two PRs that need to be merged for it not to fail on these tests. These are #24 and #26. I will wait a few days before I merge them, to allow others to comment on it.

[cvanelteren]

I am merging this with devel. After locally testing the open PRs it all looks good for now.