Full backup may slowdown during leader eviction

## Bug Report

Please answer these questions before submitting your issue. Thanks!

### 1. Minimal reproduce step (Required)
After we implement store level backup in v8.5.

When backup starts during leader eviction, a store backup may incorrectly consider itself finished and return early.

The backup will only restart in the next round. If a single round takes a long time, this behavior significantly extends the backup retention window and can effectively double the total backup time.


### 2. What did you expect to see? (Required)
All stores participate in the backup process and continue backing up data as expected during the backup round.
### 3. What did you see instead (Required)
Some stores did not perform any backup work during this round and returned early, even though the backup was still in progress.
### 4. What is your TiDB version? (Required)
v8.5

The issue names no source files, tests, or entry points. Start by locating the v8.5 store-level backup and leader-eviction paths, then trace their interaction during a backup round. Done means every store continues participating during leader eviction instead of returning early.

Written by the indexing model from the issue text.