plan: use Column.UniqueID in conditionChecker of ranger

[eurekaka]

What problem does this PR solve?

Fix #8223

What is changed and how it works?

• Compare UniqueID of Column in ranger instead of ColName, then even if the column name is incorrect, correct index scan can be chosen;
• Keep correct column name in projection elimination.

Both approaches can fix the bug independently, but seems we'd better to have them all.

We have to keep the original column name restore logic in resolveColumnAndReplace in order to maintain the compatibility with MySQL for this query:

mysql> select distinct a, (select b from tb where b = ta.a) from ta;
+---+-----------------------------------+
| a | (select b from tb where b = ta.a) |
+---+-----------------------------------+
| 1 |                                 1 |
+---+-----------------------------------+
1 row in set (0.00 sec)

if we remove line 54 and 56 from TiDB source code:

 51 func resolveColumnAndReplace(origin *expression.Column, replace map[string]*expression.Column) {
 52     dst := replace[string(origin.HashCode(nil))]
 53     if dst != nil {
 54         colName, retType := origin.ColName, origin.RetType
 55         *origin = *dst
 56         origin.ColName, origin.RetType = colName, retType
 57     }
 58 }

we would get result like:

mysql> select distinct a, (select b from tb where b = ta.a) from ta;
+---+---+
| a | b |
+---+---+
| 1 | 1 |
+---+---+
1 row in set (0.00 sec)

Check List

Tests

• Unit test

Side effects

N/A

Related changes

• Need to cherry-pick to the release branch

[eurekaka]

/run-all-tests

[eurekaka]

/run-all-tests

[XuHuaiyu]

LGTM

[winoros]

lgtm