Backup: control the costs of CPU resources on checksum operation.

[3pointer]

Feature Request

Describe your feature request related problem:

When backup lots of tables(60k+), the default checksum-concurrency(4) might cost too much CPU resources.
When backup a big table, the default checksum-concurrency(4) might not proper to finish the checksum.
so we need to discuss a proper way to control the checksum operation in different scenarios.

Describe the feature you'd like:

1. computing the checksum-concurrency in different scenarios.
2. checksum table by sampling.
3. pipeline backup table and checksum.
4. checksum offline.

Describe alternatives you've considered:

Teachability, Documentation, Adoption, Migration Strategy: