Use size of filtered column to reserve memory for filter when result_size_hint < 0 (#6746)

[ti-chi-bot]

This is an automated cherry-pick of #6746

What problem does this PR solve?

Issue Number: close #6730

Problem Summary:
See #6730

What is changed and how it works?

If result_size_hint < 0, filter makes reserve using size of filtered column, not source column.

Check List

Tests

• Unit test
• Integration test
• Manual test (add detailed scripts or steps below)
• No code

/// c is a string that is not equal to every b in table.
mysql> create table oom(a bigint, b varchar(200), c char(200));
Query OK, 0 rows affected (0.09 sec)

mysql> select count(*) from oom;
+----------+
| count(*) |
+----------+
|    60000 |
+----------+
1 row in set (0.02 sec)

mysql> alter table oom set tiflash replica 1;
Query OK, 0 rows affected (0.09 sec)

mysql> explain select * from oom o1 where (a,b,c) not in (select a,b,c from oom o2);
+--------------------------------------+----------+--------------+---------------+-------------------------------------------------------------------------------------------------------------------------+
| id                                   | estRows  | task         | access object | operator info                                                                                                           |
+--------------------------------------+----------+--------------+---------------+-------------------------------------------------------------------------------------------------------------------------+
| TableReader_27                       | 48000.00 | root         |               | MppVersion: 1, data:ExchangeSender_26                                                                                   |
| └─ExchangeSender_26                  | 48000.00 | mpp[tiflash] |               | ExchangeType: PassThrough                                                                                               |
|   └─HashJoin_25                      | 48000.00 | mpp[tiflash] |               | CARTESIAN anti semi join, other cond:eq(test.oom.a, test.oom.a), eq(test.oom.b, test.oom.b), eq(test.oom.c, test.oom.c) |
|     ├─ExchangeReceiver_14(Build)     | 60000.00 | mpp[tiflash] |               |                                                                                                                         |
|     │ └─ExchangeSender_13            | 60000.00 | mpp[tiflash] |               | ExchangeType: Broadcast                                                                                                 |
|     │   └─TableFullScan_12           | 60000.00 | mpp[tiflash] | table:o2      | keep order:false                                                                                                        |
|     └─TableFullScan_11(Probe)        | 60000.00 | mpp[tiflash] | table:o1      | keep order:false                                                                                                        |
+--------------------------------------+----------+--------------+---------------+-------------------------------------------------------------------------------------------------------------------------+
7 rows in set (0.00 sec)

In master, this sql can reach 20GiB+ memory and OOM on my laptop.
After this patch, this sql can run successfully and the max memory is about 500MiB.

Side effects

• Performance regression: Consumes more CPU
• Performance regression: Consumes more Memory
• Breaking backward compatibility

Documentation

• Affects user behaviors
• Contains syntax changes
• Contains variable changes
• Contains experimental features
• Changes MySQL compatibility

Release note

Fix the issue that CARTESIAN semi join may use larger memory than expected.

[ti-chi-bot]

[REVIEW NOTIFICATION]

This pull request has been approved by:

• LittleFall
• gengliqi

To complete the pull request process, please ask the reviewers in the list to review by filling /cc @reviewer in the comment.
After your PR has acquired the required number of LGTMs, you can assign this pull request to the committer in the list by filling /assign @committer in the comment to help you merge this pull request.

The full list of commands accepted by this bot can be found here.

Reviewer can indicate their review by submitting an approval review.
Reviewer can cancel approval by submitting a request changes review.

[ti-chi-bot]

@ti-chi-bot: This cherry pick PR is for a release branch and has not yet been approved by release team.
Adding the do-not-merge/cherry-pick-not-approved label.

To merge this cherry pick, it must first be approved by the collaborators.

AFTER it has been approved by collaborators, please ping the release team in a comment to request a cherry pick review.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

[ti-chi-bot]

This pull request is closed because its related version has closed automatic cherry-picking.
If it's still needed, you can reopen it or just regenerate it using bot,
see:

https://prow.tidb.net/command-help#cherrypick
https://book.prow.tidb.net/#/plugins/cherrypicker