support numpy scalar indexing

[mm65x]

Proposed changes

#2710

indexing with numpy integer scalars (np.int64(1), np.int32(2), etc.) throws
"Cannot index mlx array using the given type" because the indexing code only
checks for Python's built-in int via nb::isinstance<nb::int_>. numpy
integer scalars aren't int subclasses but they do implement __index__
(PEP 357).

the fix replaces all nb::isinstance<nb::int_> checks in the indexing path
with a helper that uses PyIndex_Check, which matches any object implementing
__index__. this covers direct indexing (x[np.int64(1)]), slicing
(x[np.int64(1):]), tuple indexing (x[np.int64(0), np.int64(1)]), and
assignment through all the same paths.

Checklist

• I have read the CONTRIBUTING document
• I have run pre-commit run --all-files to format my code / installed pre-commit prior to committing changes
• I have added tests that prove my fix is effective or that my feature works
• I have updated the necessary documentation (if needed)

[angeloskath]

There is already a PR implementing the exact same fix #3210, feel free to comment there if there is something that isn't supported or can be improved.

It has also been several PRs that already had PRs submitted with the exact same fix (a lot of which I merged because they were small). Please do a search and contribute to the existing PRs before submitting a new one.