Query in transaction may return rows with same unique index column value

Bug Report

Please answer these questions before submitting your issue. Thanks!

1. Minimal reproduce step (Required)

drop table if exists t;
create table t (a int, b int, primary key (a, b) /*T![clustered_index] nonclustered */);
insert into t values (1, 10);
begin optimistic;
insert into t values (1, 10);
select * from t;
commit;

2. What did you expect to see? (Required)

Only one row is returned by the select statement.

3. What did you see instead (Required)

mysql> /* t */ select * from t; -- (1, 10), (1, 10)
+---+----+
| a | b  |
+---+----+
| 1 | 10 |
| 1 | 10 |
+---+----+
2 rows in set (0.00 sec)

mysql> /* t */ commit;
ERROR 1062 (23000): Duplicate entry '1-10' for key 'PRIMARY'

4. What is your TiDB version? (Required)

Release Version: v5.0.0-43-g41871e0c8 Edition: Community Git Commit Hash: 41871e0c8e5e4b0601dd94d09342bc0317cbb8cb Git Branch: release-5.0 UTC Build Time: 2021-04-20 09:24:52 GoVersion: go1.13 Race Enabled: false TiKV Min Version: v3.0.0-60965b006877ca7234adaced7890d7b029ed1306 Check Table Before Drop: false