Slow queries might meet failure with `Can not find disaggregated task`

Enhancement

In the internal testing disagg-tiflash-workload.ch-partition-hash-run. We meet a query failure as follow

        Error 1105 (HY000): other error for mpp stream: Code: 11004, e.displayText() = DB::Exception: Check snap != nullptr failed: Can not find disaggregated task, task_id=DisTaskId<MPP<gather_id:1, query_ts:1748028447960569885, local_query_id:21
326, server_id:1629, start_ts:458235169457307650,task_id:15>,executor=TableFullScan_81> (from s1008_t5124_12772_3_22445363382982345_177150), e.what() = DB::Exception,

This is because the profiles.default.disagg_task_snapshot_timeout is 5 minute. Under disagg arch, if a slow query that TiFlash compute node can not fetch all mem-tables from the TiFlash write node in 5 minutes, then the query will meet this error and fail.

https://github.com/pingcap/tiflash/blob/b964bf581be2968e1f7ac8b07ee2cf1e7d6eb546/dbms/src/Interpreters/Settings.h#L256
https://github.com/pingcap/tiflash/blob/b964bf581be2968e1f7ac8b07ee2cf1e7d6eb546/dbms/src/Core/Defines.h#L50-L53

Related logging:
0530_endless_7865705.tar.gz

Use tar xvf 0530_endless_7865705.tar.gz for decompress

Start with dbms/src/Interpreters/Settings.h at the disagg_task_snapshot_timeout setting and dbms/src/Core/Defines.h at the related timeout definition. Review the attached workload log and the disaggregated task snapshot handling to understand why a slow query loses its task. Done means slow queries no longer fail solely because fetching mem-tables exceeds the current timeout.