SubTracker: Convert static types (PyTypeObject) to heap types via PyType_FromModuleAndSpec #32747

Part of #32451 and #31913. This is intended to be a tracking issue for a series of PRs converting NumPy's static types to heap types.

NumPy currently defines its types (ndarray, dtype, ufunc, scalars, iterators, etc.) as static PyTypeObjects. Static types are shared by every interpreter in the process, including their tp_dict and reference counts, so they can't be isolated per interpreter. The CPython Isolating Extension Modules guide recommends replacing them with heap types created by PyType_FromModuleAndSpec.

This issue depends on multi-phase init (#29021) and generally follows the Changing Static Types to Heap Types section of the guide:

Converting to heap types is also a prerequisite for building against the Limited API, where static PyTypeObject definitions are not possible, see #31913.