kv client resolves lock immediately after ticdc is up

[overvenus]

Bug Report

Please answer these questions before submitting your issue. Thanks!

1. What did you do? If possible, provide a recipe for reproducing the error.

https://github.com/pingcap/ticdc/blob/b6451b6481538a92ddfdfe6de5a4901789e0b159/cdc/kv/region_worker.go#L293-L315

$ less cdc.log
[2021/06/29 15:40:31.453 +08:00] [WARN] [region_worker.go:303] 
["region not receiving resolved event from tikv or resolved ts is not pushing for too long time, try to resolve lock"] 
[regionID=1031073] 
[span="[7480000000000023ffe95f728000000000ff0021960000000000fa, 7480000000000023ffe95f728000000000ff00b70e0000000000fa)"] 
[duration=74h26m11.05s] [lastEvent=821.902081ms] [resolvedTs=425901283045015552]

$ cdc cli changefeed list
  {
    "id": "bl1",
    "summary": {
      "state": "normal",
      "tso": 425901283045015552,
      "checkpoint": "2021-06-26 13:14:19.858",
      "error": {
        "addr": "172.16.5.33:47912",
        "code": "CDC:ErrOwnerUnknown",
        "message": "rpc error: code = DeadlineExceeded desc = context deadline exceeded"
      }
    }
  }

2. What did you expect to see?

Do not resolve lock so quickly.

3. What did you see instead?

4. Versions of the cluster

Release-5.0 branch b6451b6

[overvenus]

https://github.com/pingcap/ticdc/blob/3c0be0d4c0e36c62971f5e57364223d1c8c70abb/cdc/kv/region_worker.go#L580-L597

It is possible to send checkpoint ts to rtsManager at line 590 and this leads to resolve lock when checkpoint ts is 20s before (the current resolve lock threshold).