chore: rm support for Python 3.9 + bump Numba

[Saransh-cpp]

Description

Fixes #600
Fixes #641

Removes Python 3.9 support:

• be explicit about zip being strict
• update typing syntax
• type ignore numpy ufunc calls for object and sympy vectors
• fix getitem signature for numpy vectors and coordinate classes
• explicit type ignores for momentum numpy classes

Bumps the minimum Numba version (supported from Python 3.9):

• remove local nan_to_num definition

Checklist

• Have you followed the guidelines in our Contributing document?
• Have you checked to ensure there aren't any other open Pull Requests for the required change?
• Does your submission pass pre-commit? ($ pre-commit run --all-files or $ nox -s lint)
• Does your submission pass tests? ($ pytest or $ nox -s tests)
• Does the documentation build with your changes? ($ cd docs; make clean; make html or $ nox -s docs)
• Does your submission pass the doctests? ($ pytest --doctest-plus src/vector/ or $ nox -s doctests)

Before Merging

• Summarize the commit messages into a brief review of the Pull request.

[Saransh-cpp]

I added this to Numba a while back

[Saransh-cpp]

This also implements __getitem__ for NumPy coordinate types. Updating the type signature allowed me to remove a few type: ignore[arg-type] comments in the function body.

[Saransh-cpp]

Updated overloads as per the updates to the actual function (see above)

[Saransh-cpp]

The type: ignore[misc] comments were already present in VectorNumpy*D definitions, but now they are also required for Momentum classes.

[Saransh-cpp]

@henryiii, @pfackeldey, could you please take a look whenever you are free (after the holidays 🎄)? Thanks!

[ikrommyd]

I was actually just looking into making a PR to drop 3.9 so it's good to see that this was already done. I was gonna add 3.14 testing though since I noticed that the ci matrix does not test against 3.14. Do you wanna do it in this PR or should I open a separate one?

[ikrommyd]

You would need to change a test for 3.14 support. I'm attempting to run the vector tests with 3.14 here scikit-hep/integration-tests#14 and I'm hitting this.

     def test_VectorObjectType():
        # These tests verify that the reference counts for Python objects touched in
        # the lowered Numba code do not increase or decrease with the number of times
        # the function is run.
    
        @numba.njit
        def zero(obj):
            return None
    
        @numba.njit
        def one(obj):
            return obj
    
        @numba.njit
        def two(obj):
            return obj, obj
    
        obj = vector.obj(x=1, y=2)
>       assert (sys.getrefcount(obj), sys.getrefcount(obj.azimuthal)) == (2, 2)
E       assert (1, 2) == (2, 2)
E         
E         At index 0 diff: 1 != 2
E         
E         Full diff:
E           (
E         -     2,
E         ?     ^
E         +     1,
E         ?     ^
E               2,
E           )

[Saransh-cpp]

Ooh, I think we somehow missed that, thanks, @ikrommyd! I can quickly create a new PR to fix this.

[ikrommyd]

Ooh, I think we somehow missed that, thanks, @ikrommyd! I can quickly create a new PR to fix this.

Thanks! If you don't have time and want me to do it, let me know. For awkward, we treated numba refcount differences like this in 3.14: https://github.com/scikit-hep/awkward/pull/3757/changes