This is an automated cherry-pick of #39887

Signed-off-by: hillium yujuncen@pingcap.com

What problem does this PR solve?

Issue Number: close #39620 (again!)

Problem Summary:
This PR make a saner config for ticking timeout, and make some sentinels for prevent us doing insane things, they are:

• Uploading global checkpoint would check the old checkpoint then do the upload. (For now, we are not sure how to compare encoded bytes, and for now the keys are operated by exactly one node (elected by etcd), so the etcd conditional transaction isn't used.)
• The tick timeout now is tickDuration, so the tick won't be skipped due to blocking.
• The threshold now is 4 mins when subscription is available, 108s when not. So once we lost some subscriptions, the probability of breaking our RPO would be less.
• The tick is now split into two parts: optional part and important part. The former includes updating subscription and polling the checkpoints, which could be skipped and won't ruin things. The latter includes updating checkpoint and GC safepoint(not added yet), etc..

What is changed and how it works?

Check List

Tests

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

Release note

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

Fixed a bug that may cause log backup checkpoint lag increases to infinity when doing rolling update.