Bump min Python to 3.9

[flying-sheep]

• Closes Drop python 3.8 support #1116
• Tests added
• Release note added (or unnecessary)

Currently, this adds all checks I think are very useful, minus the following that require more substantive changes:

• B028/no-explicit-stacklevel: we should do this for more useful warning stack locations
• FBT/flake8-boolean-trap: very useful for a future API update
• PT/flake8-pytest-style very useful, but needs a lot of changes
• PD/pandas-vet

The following stylistic rulesets are useful enough to go for but also need quite some changes

• D/pydocstyle
• RET/[return style] Clarifies early returns (if thing: return\nelse: ... → if thing: return\...)

Instead of FA we just add from __future__ import annotations automatically everywhere.

The idea is that before the 0.10 release, we

1. git reset --hard <first commit in this PR> and then git rebase main this branch
2. re-create the “pre-commit autofixes” commit or let pre-commit handle that
3. maybe fix up some things that got less pretty, e.g.

• wrap braces arount this type:

  anndata/anndata/_core/index.py

  Lines 41 to 47 in b746b97

  	indexer: slice
  	| np.integer
  	| int
  	| str
  	| Sequence[int | np.integer]
  	| np.ndarray
  	| pd.Index,

• manually switch from cabc.??? to from collections.abc import ???

6. profitmerge

[codecov]

Codecov Report

Merging #1117 (d801da6) into main (8c00a29) will decrease coverage by 3.30%.
The diff coverage is 62.88%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1117      +/-   ##
==========================================
- Coverage   84.97%   81.67%   -3.30%     
==========================================
  Files          36       36              
  Lines        5197     5305     +108     
==========================================
- Hits         4416     4333      -83     
- Misses        781      972     +191     

Flag	Coverage Δ
gpu-tests	?

Flags with carried forward coverage won't be shown. Click here to find out more.

Files Changed	Coverage Δ
anndata/experimental/merge.py	82.60% <27.27%> (-4.96%)	⬇️
anndata/experimental/pytorch/_annloader.py	22.22% <45.45%> (+1.60%)	⬆️
anndata/_core/file_backing.py	86.59% <50.00%> (-3.73%)	⬇️
anndata/compat/exceptiongroups.py	85.71% <50.00%> (+2.38%)	⬆️
anndata/tests/helpers.py	85.09% <50.00%> (-10.92%)	⬇️
anndata/_core/anndata.py	81.56% <54.54%> (-2.04%)	⬇️
anndata/experimental/multi_files/_anncollection.py	71.25% <56.66%> (+0.68%)	⬆️
anndata/_core/index.py	92.02% <57.89%> (-2.59%)	⬇️
anndata/_core/merge.py	81.26% <58.13%> (-12.85%)	⬇️
anndata/__init__.py	72.00% <60.00%> (-0.73%)	⬇️
... and 26 more

📢 Have feedback on the report? [Share it here](https://about.codecov.io/codecov-pr-comment-feedback/?utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=scverse).

[ivirshup]

I've merged a version of this PR closer to the set of changes from when I reviewed last week. I did include a few more of the rules you added here.

Could you open an issue for follow up?

From time spent investigating these changes:

• "ICN" I would actually like to go a little further than this rule set goes, and be able to define our own import patterns. I think this would be useful when dealing with all the libraries that define a sparse or linalg module. I would also like to have rules that make import from anndata in the tests more like how we recommend people use them, so it's easier to copy and past code out of the tests into a repl or notebook.
• "SIM", I think I'm against this one. Some things it made better, some things it made worse. I think these should be almost always better.
• For the things around typing, I'm a little curious about them. I think I'd like to have type checking actually run before doing too much here though.
• One of the rules was replacing any instance of k in x.keys() with x in k, which can be different things – especially if we aren't runtime type checking the inputs