search: Reduce failures in shuffle/reduce phase of subgraph_query_mapping_daily

Some days with larger datasets were failing in the new spark 3 instance. The failures were related to timeouts around shuffles. Apply settings we've used on other dags that had similar issues.