Cleaning up generating macros in _cursesmodule.c

[picnixz]

This is mostly a cosmetic but useful PR. Why? because it's for future maintenance. For now, all generated methods have the following signature (modulo the Py_UNUSED):

static PyObject *NAME(PyCursesWindowObject *self, PyObject *Py_UNUSED(args));

and they are stored in PyCursesWindow_methods using (PyCFunction) casts. I'm not sure whether we want to keep an explicit cast here or not but since we are also fixing UBSan failures, I think this one would count as one.

Instead of changing it now, I first suggest cleaning up the macros and then, once we also fixed Argument Clinic generation, we may then fix the UBSan failures here as well (or I can patch half of the Window's object methods since the other half is AC-generated).

I don't see when we'll need to call a method directly using its implementation and not via thePyObject_Call* API so we can also patch the macros directly in this PR.

@vstinner please tell me whether you want me to include the UBSan failure fix in this PR or not (for that reason, I will not put an issue number yet since I don't know whether it should be part of #123961 or #111178).

[picnixz]

The commit 6f84b18 is only for removing the stateful part in the names of the functions, making most of the functions named with a shorter name. There are only 3-4 occurrences and usage of a stateless check but many occurrences of a stateful one.

If you want a separate PR, I can just cherry-pick that commit though it's easier to include it in this one to reduce the number of whitespace changes.

[vstinner]

I don't see the point of these changes, they don't solve any problem. They only look like coding style changes: -1 for me.

[picnixz]

Oh I should have given more context. I wanted to add the possibility to include the Python function name and the curses function name that were involved when we raise an exception. For now, it's a generic message saying that function X raised something, but the names may not necessarily match the C function that was actually called.

What I wanted to do is essentially mimic OSError and its filename/filename2 attributes (that could help instead of parsing the message that we could change in the future). So I wanted first to cleanup the base code before continuing the work.

---

If you think that the error feature is not needed, I'll just close this PR and the refactorization issue as well.

[picnixz]

Closing in favor of #125844.