Add num_waiting_uncached_tokens load metric (#27174)

Co-authored-by: cctry <cctry@fb.com>
Co-authored-by: Lianmin Zheng <lianminzheng@gmail.com>

test/registered/unit/mem_cache/test_radix_force_miss.py

@@ -36,9 +36,13 @@ class _StubReq:
         self.last_host_node = None
         self.best_match_node = None
         self.host_hit_length = None
+        self.num_matched_prefix_tokens = 0
         self.mamba_branching_seqlen = None
         self.cache_protected_len = None
 
+    def _compute_max_prefix_len(self, input_len):
+        return max(input_len - 1, 0)
+
 
 class TestZeroMatchResult(unittest.TestCase):
     def test_zero_replaces_indices_and_nodes(self):