Update nanobind to 2.15.0

[XXXXRT666]

Proposed changes

Update nanobind from 2.13.0 to 2.15.0 for the MLX Python bindings and extension examples.

This change also:

• adds coverage for the updated nanobind integer conversion semantics, accepting objects that implement __index__ while rejecting floating-point and string inputs
• updates the stub generation patterns for nanobind 2.15, including generated imports, aliases, comparison return types, and context manager signatures
• adds non-blocking Ruff and ty checks for generated mlx.core stubs in the Linux CPU CI job

Checklist

Put an x in the boxes that apply.

• 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)

[XXXXRT666]

With the updated stubs, ty check completes without warnings, and array comparisons are correctly inferred to return an mx.array:

a = mx.array(...)
b = a == a  # correctly inferred as mx.array rather than bool | mx.array

[zcbenz]

uv check seems to be ty now?

Since type checking passes now I would like to make it mandatory, I don't care about making every type checker happy and we can drop Ruff.

[XXXXRT666]

uv check relies on the project metadata in pyproject.toml, it cannot be scoped to just the generated stubs I think. Running it at the repository root currently reports hundreds of diagnostics. A full Ruff check on the generated stubs also reports 150 errors

I dropped the Ruff step and made the targeted uvx ty check python/mlx/core check mandatory.

[zcbenz]

That makes sense, thanks for fixing the type errors!