fix: retry on RuntimeError for stale Redis connections (Fixes HONCHO-18H)

uvloop raises RuntimeError when a TCP transport is closed and redis-py's
health check PING hits the stale connection. redis-py's default retry
logic doesn't cover RuntimeError, so it propagated unhandled. Adding
RuntimeError to retry_on_error causes redis-py to reconnect instead.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

src/cache/client.py

@@ -49,6 +49,7 @@ async def init_cache() -> None:
             cache.setup(
                 settings.CACHE.URL,
                 pickle_type=PicklerType.SQLALCHEMY,
+                retry_on_error=[RuntimeError],
             )
 
         except Exception as setup_err: