Many heartbeat update operations may cause GetRegion timeout

## Feature Request

### Describe your feature request related problem
As user reports, there are about 200w regions, and the workload is uniform. so there are many region need to update if the flow changed. the qps `update_cache` is about 9k, which means 9k/s need Locks and more operator in heartbeat need`RLocks`.

### Describe the feature you'd like
- reduce the update operations
- Alleviate lock contention

### Describe alternatives you've considered
- add a switch to reduce updates when region flow changed.
- optimize heartbeat with segmented lock

### Teachability, Documentation, Adoption, Migration Strategy
- Go