streaming: switch to stream_async + iter_chunks response drain

src/lambdas/agent-runner/handler.py

@@ -187,8 +187,10 @@ def handler(event, context):
         payload=json.dumps(payload).encode(),
     )
 
-    # Consume streaming response (agent delivers to Telegram via send_message tool)
-    for chunk in response.get('response', []):
-        pass  # intentional no-op — agent handles delivery internally
+    # Drain streaming response body (agent delivers to Telegram via send_message tool)
+    body = response.get('response')
+    if body is not None:
+        for _ in body.iter_chunks():
+            pass
 
     print(f"[agent-runner] Completed session={session_id} actor={actor_id}")