Fix real-time task progress by closing WS on pubsub exit and keeping polling active

Three interconnected bugs prevented progress updates from reaching the frontend:

1. _forward_pubsub could exit silently while _handle_client_messages kept
   the WebSocket alive (responding to pings), so the client never detected
   the broken forwarding path. Replace asyncio.gather with asyncio.wait
   (FIRST_COMPLETED) so both coroutines are cancelled together.

2. Polling was stopped on WS connect with no fallback if forwarding broke.
   Now polling runs always alongside WebSocket as a safety net.

3. Redis publish failures in task_progress_publisher were logged at DEBUG
   and the broken client was reused forever. Log at WARNING and reset the
   client so the next call reconnects.

services/task_progress_publisher.py

@@ -35,8 +35,9 @@ def publish_task_progress(task_id: str, state: str, meta: dict[str, Any]) -> Non
         state: Celery state string (e.g. 'PROGRESS', 'SUCCESS').
         meta: Metadata dict (progress, phase, logs, counters, etc.).
 
-    Failures are caught and logged at DEBUG level so they never break the
-    critical task execution path.
+    Failures are caught and logged at WARNING level so they never break the
+    critical task execution path.  The Redis client is reset on failure so
+    subsequent calls can reconnect.
     """
     try:
         client = _get_redis_client()
@@ -47,4 +48,6 @@ def publish_task_progress(task_id: str, state: str, meta: dict[str, Any]) -> Non
         })
         client.publish(f"task_progress:{task_id}", payload)
     except Exception:
-        logger.debug("Failed to publish task progress for %s", task_id, exc_info=True)
+        logger.warning("Failed to publish task progress for %s", task_id, exc_info=True)
+        # Reset client so next call creates a fresh connection
+        _redis_client = None