Make PyInterpreterState.threads.count thread-safe in free-threaded builds

Use atomics

Author: mpage

Include/internal/pycore_interp.h

@@ -112,7 +112,7 @@ struct _is {
         /* The thread currently executing in the __main__ module, if any. */
         PyThreadState *main;
         /* Used in Modules/_threadmodule.c. */
-        long count;
+        Py_ssize_t count;
         /* Support for runtime thread stack size tuning.
            A value of 0 means using the platform's default stack size
            or the size specified by the THREAD_STACK_SIZE macro. */

Modules/_threadmodule.c

@@ -1225,7 +1225,7 @@ thread_run(void *boot_raw)
 
     _PyThreadState_Bind(tstate);
     PyEval_AcquireThread(tstate);
-    tstate->interp->threads.count++;
+    _Py_atomic_add_ssize(&tstate->interp->threads.count, 1);
 
     PyObject *res = PyObject_Call(boot->func, boot->args, boot->kwargs);
     if (res == NULL) {
@@ -1243,7 +1243,7 @@ thread_run(void *boot_raw)
 
     thread_bootstate_free(boot, 1);
 
-    tstate->interp->threads.count--;
+    _Py_atomic_add_ssize(&tstate->interp->threads.count, -1);
     PyThreadState_Clear(tstate);
     _PyThreadState_DeleteCurrent(tstate);
 
@@ -1520,7 +1520,7 @@ static PyObject *
 thread__count(PyObject *self, PyObject *Py_UNUSED(ignored))
 {
     PyInterpreterState *interp = _PyInterpreterState_GET();
-    return PyLong_FromLong(interp->threads.count);
+    return PyLong_FromLong(_Py_atomic_load_ssize(&interp->threads.count));
 }
 
 PyDoc_STRVAR(_count_doc,