gh-117657: Fix data races report by TSAN unicode-hash

[corona10]

• Issue: Make TSAN tests pass with the GIL disabled in free-threaded builds #117657

[corona10]

@colesbury @mpage
From my local clang-18 and GHA do not report race condition issues for unicode-hash, so IMO, we can remove the function from the list until it is detected in the future.

[colesbury]

I think there is still a data race, even if it's only reported occasionally. We read and write to _PyUnicode_HASH(self) without atomics -- I think we should use relaxed atomic load/stores.

cpython/Objects/unicodeobject.c

Lines 11391 to 11396 in 90ec19f

	if (_PyUnicode_HASH(self) != -1)
	return _PyUnicode_HASH(self);
	x = _Py_HashBytes(PyUnicode_DATA(self),
	PyUnicode_GET_LENGTH(self) * PyUnicode_KIND(self));
	_PyUnicode_HASH(self) = x;

[corona10]

I think there is still a data race, even if it's only reported occasionally. We read and write to _PyUnicode_HASH(self) without atomics -- I think we should use relaxed atomic load/stores.

Thank you for the feedback, I will take a look.

[colesbury]

LGTM

[miss-islington-app]

Thanks @corona10 for the PR 🌮🎉.. I'm working now to backport this PR to: 3.13.
🐍🍒⛏🤖

[bedevere-app]

GH-119963 is a backport of this pull request to the 3.13 branch.