processor: fix task status flushed too many before table is initialized (#1190)

[ti-srebot]

cherry-pick #1190 to release-4.0

---

What problem does this PR solve?

Fix https://github.com/pingcap/ticdc/issues/1188. This is a hotfix for current codebase, we should process with the Dirty field of task status carefully.

What is changed and how it works?

change flush check condition

Check List

Tests

• Unit test

• Integration test

• Manual test (add detailed scripts or steps below)

  • hack puller p.tsTracker.Frontier() to force no forward to simulate long time incremental scan.
  • watch the task status key version in etcd doesn't change.

Release note

• Fix task status flushed to etcd too many before table is initialized

[ti-srebot]

/run-all-tests

[codecov-io]

Codecov Report

Merging #1191 (cfd19ae) into release-4.0 (ba68bc1) will decrease coverage by 0.0798%.
The diff coverage is 0.0000%.

@@                 Coverage Diff                 @@
##           release-4.0      #1191        +/-   ##
===================================================
- Coverage      39.8383%   39.7584%   -0.0799%     
===================================================
  Files              112        112                
  Lines            11750      11756         +6     
===================================================
- Hits              4681       4674         -7     
- Misses            6597       6610        +13     
  Partials           472        472                

[amyangfei]

LGTM

[liuzix]

LGTM