This is an automated cherry-pick of #41036

What problem does this PR solve?

Issue Number: close #41047

Problem Summary:
There are five kinds of workers in IndexMergeReaderExecutor:

1. partialIndexWorker/partialTableWorker: get handles from rows, then send handles to processWorker.
2. processWorker: do Union or Intersection of handles., then send handles to tableScanWorker. Also it send task to main goroutine using resultCh.
3. tableScanWorker: fetch row from tikv using handles, then send rows to main goroutine
4. partTblIntersectionWorker: For dynamic partition table, for each partition, start a goroutine to do intersection. This is to speed up intersection processing.
5. main goroutine: wait for tableScanWorker to get rows.

Why panic:

1. processWorker finish, so it close resultCh. code
2. partialIndexWorker got error, then try to send error using resultCh. code
3. got write to closed channel panic.

What is changed and how it works?

1. partialIndexWorker/partialTableWorker should not use resultCh to send error. Instead should use fetchCh, which is used to send task to processWorker.
2. processWorker should check err in task, which is received from fetchCh. If got error, just stop processing.
3. add handleIntersectionProcessWorkerPanic(), which is similar to handleLoopFetcherPanic. The latter will close(resultCh) when return.

Check List

Tests

• Unit test
• Integration test
• Manual test (add detailed scripts or steps below)
  1. load data using lighning, too large to upload
  2. run sql as follows, and tidb doesn't crash when SQL is canceled because oom.

use s11_indexmerge_test;

set global tidb_server_memory_limit="500MB";
explain analyze select /*+ use_index_merge(tp2) */ * from tp2 where a < 40000000 or b < 40000000;
explain analyze select /*+ use_index_merge(tp2, ia, ib) */ * from tp2 where a < 40000000 and b < 40000000;

set tidb_mem_quota_query = 1000;
explain analyze select /*+ use_index_merge(tp2) */ * from tp2 where a < 40000000 or b < 40000000;
explain analyze select /*+ use_index_merge(tp2, ia, ib) */ * from tp2 where a < 40000000 and b < 40000000;

set tidb_mem_quota_query = 10<<20;
explain analyze select /*+ use_index_merge(tp2) */ * from tp2 where a < 40000000 or b < 40000000;
explain analyze select /*+ use_index_merge(tp2, ia, ib) */ * from tp2 where a < 40000000 and b < 40000000;

set tidb_mem_quota_query = 500<<20;
explain analyze select /*+ use_index_merge(tp2) */ * from tp2 where a < 40000000 or b < 40000000;
explain analyze select /*+ use_index_merge(tp2, ia, ib) */ * from tp2 where a < 40000000 and b < 40000000;
                                                                                                                                                                                                                                                            set tidb_mem_quota_query = 1<<30;
explain analyze select /*+ use_index_merge(tp2) */ * from tp2 where a < 40000000 or b < 40000000;
explain analyze select /*+ use_index_merge(tp2, ia, ib) */ * from tp2 where a < 40000000 and b < 40000000;

• No code

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

Please refer to Release Notes Language Style Guide to write a quality release note.

executor: fix IndexMerge handle panic logic