Fix data race when using shared variables (free threading) (#5494)

* Fix data race when using shared variables (free threading)

In the free threading build, there's a race between wrapper re-use and
wrapper deallocation. This can happen with a static variable accessed by
multiple threads.

Fixing this requires using some private CPython APIs: _Py_TryIncref and
_PyObject_SetMaybeWeakref. The implementations of these functions are
included until they're made available as public ("unstable") APIs.

Fixes #5489

* style: pre-commit fixes

* Avoid unused parameter

* Add missing return for default build

* Changes from review

* Assign result to local variable

* s/clang-tidy/ruff

* clang-tidy: static is redundant

* Use 'noqa: B018'

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

tests/test_thread.cpp

@@ -28,6 +28,9 @@ struct IntStruct {
     int value;
 };
 
+struct EmptyStruct {};
+EmptyStruct SharedInstance;
+
 } // namespace
 
 TEST_SUBMODULE(thread, m) {
@@ -61,6 +64,9 @@ TEST_SUBMODULE(thread, m) {
         },
         py::call_guard<py::gil_scoped_release>());
 
+    py::class_<EmptyStruct>(m, "EmptyStruct")
+        .def_readonly_static("SharedInstance", &SharedInstance);
+
     // NOTE: std::string_view also uses loader_life_support to ensure that
     // the string contents remain alive, but that's a C++ 17 feature.
 }