Incremental backup should not be taken on previously-not-ready Pod

[morimoto-cybozu]

Describe the bug

When taking an incremental backup, we should not use a Pod which was not ready at the time of the previous backup.
The current BackupManager.ChoosePod() avoids to use a Pod which has the same index and a different UUID to the recorded ones, but it may choose an inappropriate Pod as an alternative.

Environments

• Version: v0.18.1
• OS:

To Reproduce

Expected behavior

An incremental backup is taken on a Pod which has sufficient binlogs.
If there's no available Pod to take binlogs, incremental backup should be skipped.

Additional context