Use free-threaded atomic wrapper in deque_len, too

Modules/_collectionsmodule.c

@@ -1214,7 +1214,7 @@ deque_contains(dequeobject *deque, PyObject *v)
 static Py_ssize_t
 deque_len(dequeobject *deque)
 {
-    return _Py_atomic_load_ssize(&((PyVarObject *)deque)->ob_size);
+    return FT_ATOMIC_LOAD_SSIZE(((PyVarObject *)deque)->ob_size);
 }
 
 /*[clinic input]