safepoint: support to set safepoints for services

[shafreeck]

Signed-off-by: Shafreeck Sea [email protected]

What problem does this PR solve?

When a TiKV cluster has multiple components rely on the GC safe point, they should have agreement of value of safe point which determines what data would be deleted.

For example, TiCDC captures the data change event after the safe point, if the safe point advances too quickly, TiCDC can not handle the case.

This PR relies on pingcap/kvproto#603

• TODO: Merge pd: add UpdateServiceGCSafePoint method pingcap/kvproto#603 and update go.mod

What is changed and how it works?

• Add UpdateServiceGCSafePoint to update the safe point for service and return the minimum value across all the services

Release note

Check List

Tests

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

Code changes

• Has persistent data change

[shafreeck]

@disksing @nolouch The feature is reimplemented and please review it again.

The original API and data are left unchanged. A new API called UpdateServiceGCSafePoint is added to coordinate the GC safe point value, it updates the safe point for service and returns the minimum value in the cluster. For TiDB, it can call UpdateGCSafePoint with the minimum value to trigger the distributed GC job or do the job locally.

[shafreeck]

@nolouch @rleungx PTAL

[sre-bot]

/run-all-tests