Add a CallBack interface that allows users to customize the behavior when requesting region data fails

[qidi1]

Feature Request

Now the cache invalid envet is all handle by client-java. We want add a CallBack interface that allows users to customize the behavior when requesting region data fails.
In TiSpark, a Job execution may involve multiple machines, and the regionMananger on each machine may have its own invalid region. Through CallBack, we can collect the invalid regions from all machines and perform invalid processing on each machine to refresh them. This reduces the probability that a subsequent Job will encounter a region failure.

[github-actions]

This issue is stale because it has been open 30 days with no activity.