Add internal API for fast module access from heap type methods

[erlend-aasland]

See topic on Discourse.

For CPython internal usage, we've got _PyModule_GetState, which is a fast variant of PyModule_GetState, the module check in the latter is simply an assert in the former.

For PyType_GetModuleState, there are three ifs (two of them implicitly in PyType_GetModule):

1. check that the given type is a heap type
2. check that the given type has an associated module
3. check that the result of PyType_GetModule is not NULL

For stdlib core extension modules, all of these conditions are always true (AFAIK). With a fast static inlined variant, for example _PyType_GetModuleState, with a fast variant of PyType_GetModule inlined, where all three conditions are assert()ed, we can speed up a heap type methods that need to access module state.

Linked PRs

• gh-101476: Add _PyType_GetModuleState #101477
• gh-101476: Use _PyType_GetModuleState where applicable #102188

[gpshead]

relevant discuss thread: https://discuss.python.org/t/a-fast-variant-of-pytype-getmodulestate/23377

i left a question on the PR regarding the API name (ALL_CAPS or not for in the API name to be similar to other fast unchecked APIs).