hotfix: address metaclass definition in PythranBuildExt

[agriyakhetarpal]

Description

I'm not sure if this is the best fix for #2228 or if this is something that should be resolved in setuptools instead, but this PR is a small fix that resolves the metaclass conflict in Pythran that is coming from setuptools's new v73.0.0 release. I can close this PR if it gets resolved elsewhere!

TypeError: metaclass conflict: the metaclass of a derived class must be a (non-strict) subclass of the metaclasses of all its bases

The test suite is running on my fork here: agriyakhetarpal#1 and I have verified that the fix works by running a few of the tests locally with both setuptools>=73 and setuptools<73 (the tests do take some time to fire away, though!), and in the meantime, #2229 pins setuptools to <73 which is an easier fix to incorporate.

Changes made

• PythranBuildExtMeta derives from type(LegacyBuildExt) – which ensures that PythranBuildExtMeta is a subclass of the metaclass used by LegacyBuildExt, and
• instead of defining a new class inside the __getitem__ method, it now uses the type() function to dynamically construct a new class.

Additional context

xref to some downstream issues where this was discovered:

1. SciPy v1.13.1 pyodide/pyodide#5011, Scipy build issue with setuptools 73.0 pyodide/pyodide#5018
2. BLD: build failure with new setuptools 73.0.0 scipy/scipy#21416
3. setuptools v73.0.0 conda-forge/setuptools-feedstock#356

[agriyakhetarpal]

The offending commit in setuptools (between version 72.0.0 and version 73.0.0) seems to be f9398b1f. I can confirm that the fix I propose through this PR works – I verified by testing a portion of the tests: python -m pytest pythran/tests/test_list.py.

[agriyakhetarpal]

Closing since this is not needed anymore: https://github.com/pypa/setuptools/releases/tag/v73.0.1. A release should be up on PyPI soon.