What problem does this PR solve?

Reduce object allocation by reuse the chunk column.
Close #28340

Problem Summary:

For every query, we will create the chunk...
This is unnecessary, after the chunk data is drained by the MySQL client, nobody will use the chunk again, so it should be safe to reuse the chunk.

What is changed and how it works?

Many changes... maybe I should split the PR to the lib and the use of the lib?

• Add a chunkAlloc for each client connection, allocate chunks from it, and reset for reuse after chunk sent to the client
• Add a NewChunkFromAllocator method to the ResultSet interface
• Update of the benchmark tests to use the chunk allocator
• Implementation of the chunk Allocator

For the chunk Allocator implementation, I decouple the chunk object into chunk column objects for reuse, because the Chunk differs a lot, while the chunk column are fix sized.

Allocating chunk will try to take a empty Chunk object from the free list, and fill the columns from the columnAlloc,
and after reset, allocated chunks will turn into free with their columns collected into the poolColumnAllocator:

type allocator struct {
	allocated   []*Chunk
	free        []*Chunk
	columnAlloc poolColumnAllocator
}

Check List

Tests

• Unit test
• Integration test
• Manual test (add detailed scripts or steps below)

cd session;
go test -bench BenchmarkPreparedPointGet -run XXX -benchtime 30s -benchmem -cpuprofile cpu.profile
BenchmarkPreparedPointGet-16    	 4350782	      8383 ns/op	    5262 B/op	      77 allocs/op
BenchmarkPreparedPointGet-16    	 4218360	      8486 ns/op	    5486 B/op	      83 allocs/op

• 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

Reduce object allocation by reuse chunk column