*: fix executing shard DDL with mismatched request (#336)

[sre-bot]

cherry-pick #336 to release-1.0

---

What problem does this PR solve?

1. DM-worker does not verify the ExecDDLRequest from the DM-master
2. DM-master may send duplicate ExecDDLRequest to DM-worker (in order to support re-fetching DDL info from DM-worker after DM-master restarted)

example:

1. execute ADD INDEX idx1
2. duplicate ADD INDEX idx1 request sent
3. ADD COLUMN c1 synced in DM-worker
4. duplicate request used, ADD COLUMN c1 executed, but ADD COLUMN c1 in other DM-workers not synced yet.

ref: https://internal.pingcap.net/jira/browse/ONCALL-745

What is changed and how it works?

1. do not re-send DDL execution request in most scenarios
   1. ignore the invalid connection for ADD INDEX
   2. decrease DDL DB execution timeout, increase DDL gRPC execution request
2. ignore duplicate DDL execution request
   • verify DDL if possible

Check List

Tests

• Unit test
• Integration test

Related changes

• Need to cherry-pick to the release branch
• Need to be included in the release note

[codecov]

Codecov Report

Merging #338 into release-1.0 will not change coverage.
The diff coverage is n/a.

@@              Coverage Diff              @@
##           release-1.0      #338   +/-   ##
=============================================
  Coverage       60.147%   60.147%           
=============================================
  Files              135       135           
  Lines            15236     15236           
=============================================
  Hits              9164      9164           
  Misses            5173      5173           
  Partials           899       899

[WangXiangUSTC]

LGTM

[amyangfei]

LGTM