executor: fix deadloop when exec replace with generated column

[amyangfei]

What problem does this PR solve?

TiDB will run into dead loop and exhausts large memory very soon if we execute replace into values (...) on a generated column. The bug can be reproduced as following:

mysql> create table t3 (id int, id_gen int as (`id`+1), unique key id_gen(`id_gen`));
Query OK, 0 rows affected (0.04 sec)

mysql> insert into t3 (id) values (1);
Query OK, 1 row affected (0.00 sec)

mysql> replace into t3 (id) values (1);

The root cause of this is we fetched key value from store for duplicated kv checks, but did not apply generated column rule.

What is changed and how it works?

Force remove duplicated key from ReplaceExec.dupKVs.

This pr is not a totally solution to this problem, it solves the deadloop, but a replace into operation having duplicated unique key conflict on a generated column still fail now.

I didn't find a suit method to calculate generated column value without building a logic planner, if anyone points out a workaround way, I will have a try to make this pr better.

Check List

Tests

• Unit test
• Integration test

Related changes

• Need to cherry-pick to the release branch

[codecov]

Codecov Report

Merging #10357 into master will decrease coverage by 0.0102%.
The diff coverage is 100%.

@@               Coverage Diff                @@
##             master     #10357        +/-   ##
================================================
- Coverage   77.6897%   77.6794%   -0.0103%     
================================================
  Files           411        411                
  Lines         85436      85446        +10     
================================================
- Hits          66375      66374         -1     
- Misses        14101      14111        +10     
- Partials       4960       4961         +1

[codecov]

Codecov Report

Merging #10357 into master will decrease coverage by 0.0266%.
The diff coverage is 100%.

@@               Coverage Diff               @@
##             master    #10357        +/-   ##
===============================================
- Coverage   77.6897%   77.663%   -0.0267%     
===============================================
  Files           411       411                
  Lines         85436     85446        +10     
===============================================
- Hits          66375     66360        -15     
- Misses        14101     14119        +18     
- Partials       4960      4967         +7

[amyangfei]

/run-all-tests

[jackysp]

I filed this PR #10370 to fix it. Not fully tested.