bpo-42536: GC track recycled tuples (GH-23623)

Several built-in and standard library types now ensure that their internal result tuples are always tracked by the garbage collector:

- collections.OrderedDict.items
- dict.items
- enumerate
- functools.reduce
- itertools.combinations
- itertools.combinations_with_replacement
- itertools.permutations
- itertools.product
- itertools.zip_longest
- zip

Previously, they could have become untracked by a prior garbage collection.
diff --git a/Lib/test/test_dict.py b/Lib/test/test_dict.py
index 9ff8b7d..4b31cdc 100644
--- a/Lib/test/test_dict.py
+++ b/Lib/test/test_dict.py
@@ -1452,6 +1452,25 @@ def items(self):
         d = CustomReversedDict(pairs)
         self.assertEqual(pairs[::-1], list(dict(d).items()))
 
+    @support.cpython_only
+    def test_dict_items_result_gc(self):
+        # bpo-42536: dict.items's tuple-reuse speed trick breaks the GC's
+        # assumptions about what can be untracked. Make sure we re-track result
+        # tuples whenever we reuse them.
+        it = iter({None: []}.items())
+        gc.collect()
+        # That GC collection probably untracked the recycled internal result
+        # tuple, which is initialized to (None, None). Make sure it's re-tracked
+        # when it's mutated and returned from __next__:
+        self.assertTrue(gc.is_tracked(next(it)))
+
+    @support.cpython_only
+    def test_dict_items_result_gc(self):
+        # Same as test_dict_items_result_gc above, but reversed.
+        it = reversed({None: []}.items())
+        gc.collect()
+        self.assertTrue(gc.is_tracked(next(it)))
+
 
 class CAPITest(unittest.TestCase):