[Bug] Dedicated Compaction for write-only table cann't clean up small files after running for a long time

### Search before asking

- [X] I searched in the [issues](https://github.com/apache/incubator-paimon/issues) and found nothing similar.

### Paimon version

0.6

### Compute Engine

flink 1.16

### Minimal reproduce step

1. start a flink job for writing to paimon table with mode "write-only" with checkpoint as 5s
2. start a Dedicated Compaction flink to to compact paimon table
3. after running over 1 days
4. I found some small files yesterday still exist

### What doesn't meet your expectations?

When writing and compact job runs for more than 2 days， I found small files with timestamp as "2023-12-28 19:43:36" still exist in 2023-12-29。

### Anything else?

_No response_

### Are you willing to submit a PR?

- [ ] I'm willing to submit a PR!

No file or test is named. Reproduce with a Paimon write-only table, Flink 1.16, 5-second checkpoints, and a dedicated compaction job running for more than two days; then trace why small files from the previous day remain. Done means the compaction job cleans up the reported old small files under this workload.

Written by the indexing model from the issue text.