lightning cannot exit when task cancelled and there's pending flush chunks, if we start again might cause "resource temporarily unavailable"

## Bug Report

Please answer these questions before submitting your issue. Thanks!

### 1. Minimal reproduce step (Required)

- if chunk flush is slow, so there might be many pending flush chunks add in here
https://github.com/pingcap/tidb/blob/ab4c06aac5f8cd3ffbb18ef1f6ac5f25d9d40916/br/pkg/lightning/importer/table_import.go#L779
- if they're flushed later, we will put them into a channel to trigger checkpoint save
https://github.com/pingcap/tidb/blob/ab4c06aac5f8cd3ffbb18ef1f6ac5f25d9d40916/br/pkg/lightning/importer/table_import.go#L694
- if we cancel the task, the routine which handles previous channel will exit, so the channel might be full, and block the exit
https://github.com/pingcap/tidb/blob/ab4c06aac5f8cd3ffbb18ef1f6ac5f25d9d40916/br/pkg/lightning/importer/table_import.go#L658-L667

the stack when it's stuck: [stack2.log](https://github.com/pingcap/tidb/files/11948867/stack2.log)

if we try to start another run of same table, we will meet `resource temporarily unavailable`, since some files in pebble db is locked by the un-exited lightning and we're using tablename to generate pebble db directory name.

### 2. What did you expect to see? (Required)
lightning exited
### 3. What did you see instead (Required)
keeps running
### 4. What is your TiDB version? (Required)
hash 60784a30d041479769aa6a4a351df5eeddd0d94a