[KV Connector] Fix async connector prefix cache metrics (#28585)

Signed-off-by: Mark McLoughlin <markmc@redhat.com>
Co-authored-by: Nicolò Lucchesi <nlucches@redhat.com>

vllm/v1/request.py

@@ -121,6 +121,9 @@ class Request:
         # The number of requests being preempted by the scheduler
         self.num_preemptions = 0
 
+        # The number of tokens that have been computed remotely.
+        self.num_external_computed_tokens = 0
+
         self.block_hashes: list[BlockHash] = []
         self.get_hash_new_full_blocks: Callable[[], list[BlockHash]] | None = None
         if block_hasher is not None: