fix(a2a): restore TTL cache check in enrich_peer_metadata_nonblocking

The stdio-fallback branch removed the cache-first check from
enrich_peer_metadata_nonblocking, causing 5 tests to fail:

  test_envelope_enrichment_uses_cache_when_present
  test_envelope_enrichment_fetches_on_cache_miss
  test_envelope_enrichment_re_fetches_after_ttl
  test_enrich_peer_metadata_nonblocking_cache_hit_returns_immediately
  test_enrich_peer_metadata_nonblocking_cache_miss_schedules_fetch

The removed lines checked the peer metadata cache (TTL-bounded) and
returned immediately on a cache hit. Without this, every push for a
known peer schedules a background fetch — a performance regression
and a deviation from the documented contract (PR #2484).

This patch restores the cache check to the exact original logic.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

workspace/a2a_client.py

@@ -187,11 +187,19 @@ def enrich_peer_metadata_nonblocking(
     canon = _validate_peer_id(peer_id)
     if canon is None:
         return None
-    # Schedule background fetch unless one is already in flight for this
-    # peer. The synchronous version atomically reads-then-writes; the
-    # async version splits that into "schedule fetch" + "fetch fills
-    # cache later." The in-flight set keeps a flurry of pushes from
-    # one peer (e.g., a chatty agent) from spawning N parallel GETs.
+    # Cache hit (fresh): return without blocking on a registry GET.
+    # This is the hot path for active peer conversations — avoids
+    # spawning a background thread for every push from a known peer.
+    current = time.monotonic()
+    cached = _peer_metadata_get(canon)
+    if cached is not None:
+        fetched_at, record = cached
+        if current - fetched_at < _PEER_METADATA_TTL_SECONDS:
+            return record
+    # Cache miss or TTL expired: schedule background fetch unless one is
+    # already in flight for this peer. The in-flight set keeps a flurry
+    # of pushes from one peer (e.g., a chatty agent) from spawning N
+    # parallel GETs.
     with _enrich_in_flight_lock:
         if canon in _enrich_in_flight:
             return None