Issues Observed During 4000 Keyspace Concurrency Test

## Enhancement Task
Performed a concurrency test involving 4000 keyspaces on the cluster and observed system behavior.

- **PD API:**
- CPU usage increased, but not fully saturated

- **TSO:**
- Primary frequently switched
- CPU usage increased noticeably
- **gRPC:**
- Noticeable increase in several requests: `Range`, `GetClusterInfo`, `GetMembers (PD)`, `MemberList (etcd client)`
- **PD etcd :**
- etcd disk requests significantly increased
- etcd observed repeated raft re-elections
- In `trace.go`, two keys had the largest request volume: `gcworker/saved_safe_point`, `tidb/server/info`
- TSO Encountered `save ts error`, further affecting ts
- **Unexpected HTTP responses:**
- `GetMinResolvedTs` and `GetAllRestores` not as expected.