When Lightning's duplicate-detection encountered a retryable error, it will retry without back-off

## Bug Report

Please answer these questions before submitting your issue. Thanks!

### 1. Minimal reproduce step (Required)

1. Run Lightning with duplicate detection enabled (set conflict strategy to not "none")
2. After `[detect-dupe] collect remote duplicate keys` started, perform some region leader transfer / split / merge
* Ideally this should never happen because of import-mode and pause scheduler, but both of them can be disabled.
3. Note a bunch of `process remote dupTask encounters region error, retrying` in the log.

### 2. What did you expect to see? (Required)

There should not be too many errors, as the detector should wait for the region cache to be refreshed.

### 3. What did you see instead (Required)

Lightning retried without any backoff. So even with 1800 retries it may still be using the old region → store mapping.

https://github.com/pingcap/tidb/blob/b25074f843ae77f4edfd2b9e558bd51a399ddb07/pkg/lightning/backend/local/duplicate.go#L1005-L1015

### 4. What is your TiDB version? (Required)