Tracking issue for large scale lightning import

[lance6716]

The checkbox means the PR has been cherry-picked to v7.1

• [Tracking] The "import" folder may occupy lots of disk space after import is finished tikv/tikv#14741
• lightning: always retry from needRescan when NotLeader #43079
• for large import task, lightning needs to improve region split/scatter/check #43510
• needRescan region job should not pass retryCount to new jobs  #43682
• lightning might oom when mem/cpu ratio is low #43728
• lightning might stuck at split/scatter region #43657
  • PD side issue: Lightning block at not full replicaed tikv/pd#6426
• retry on unknown RPC error from TiKV #43291
• change log level to debug of "keys within region is empty, skip doIngest" #43197
• lightning: some of lightning's grafana panels lack data #43357
• lightning: TiKV space check affects performance #43636
• lightning engines are not pipelined, what's the difference of using 6 engines or 1 engine? #43750
• lightning parallel import index engine will cause many files ingested to L0  #43756
• in parallel import, lightning's two rounds of waiting for scatter is problematic #43760
• support GOMEMLIMIT internally, see https://github.com/pingcap/tidb/blob/master/util/gctuner/memory_limit_tuner.go
• lightning oom when importing wide table on default config #43853