gh-134745: Change PyThread_allocate_lock() implementation to PyMutex

[vstinner]

PyMutex is a fast and portable lock currently only used internally by Python. This change modifies PyThread_allocate_lock to reuse PyMutex.

• Issue: Change PyThread_allocate_lock() implementation to PyMutex #134745

[vstinner]

This draft PR currently only modify Python/thread_pthread.h. I left Python/thread_nt.h unchanged.

_PyThread_at_fork_reinit() is still implemented by allocating a new lock and leaking the old lock.

[vstinner]

@colesbury: What do you think of this idea? Does it make sense? Do you see any possible issue?

If the approach makes sense, I will update my PR to modify also the Windows implementation.

[colesbury]

I think the idea is fine, but the PR title and description are confusing. threading.Lock / _thread.lock / _thread._thread.allocate_lock() already use PyMutex (all three are the same thing):

cpython/Modules/_threadmodule.c

Lines 757 to 760 in f6f4e8a

	typedef struct {
	PyObject_HEAD
	PyMutex lock;
	} lockobject;

This PR changes the implementation of the C API PyThread_allocate_lock(). We are generally using PyMutex directly instead of PyThread_allocate_lock() in new code, and we've also changed a lot of existing uses too. Using PyMutex directly is simpler because it avoids an allocation that may fail.

I think it's okay to change the implementation of PyThread_allocate_lock(). It looks like it's going to simplify the code, but there's some risk of unknown behavioral changes.

[colesbury]

This LGTM, but I think you should include a NEWS entry. One behavioral change is that previously intr_flag was ignored on Windows, but now it actually makes the acquisitions interruptible.

[vstinner]

Merged, thanks for the review @colesbury.