[Bug] remove_orphan_files may accidentally delete data in multitasking concurrent scenarios

### Search before asking

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

### Paimon version

version 1.2

### Compute Engine

Spark 3.5.6

### Minimal reproduce step

table max retained snapshot is 1
'snapshot.num-retained.max' = '1'

2026-04-25 04:37:12,116, Task 1 completed sys.compact operation and submitted snapshot 4959
2026-04-25 04:37:13,195, Task 1 completed snapshot expiration. Mark snapshot 4958 as expired and delete it
2026-04-25 04:37:37,605, Task 2 completed writing and submitted snapshot 4960
2026-04-25 04:37:38,185, Task 2 completed snapshot expiration. Mark snapshot 4959 as expired and delete it
At the same time, Task 1 started executing the remove_orphan_files, but the current snapshot 4959 was not read because 4959 had already been deleted by Task 2, resulting in usedFiles being empty, which in turn caused deleted=candidates, meaning that all files written one day ago were marked as files to be deleted and ultimately deleted.

### What doesn't meet your expectations?

Incorrect deletion of historical data

### Anything else?

_No response_

### Are you willing to submit a PR?

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

Start with the remove_orphan_files flow and the concurrent snapshot-expiration sequence described for Paimon 1.2 on Spark 3.5.6. Trace how snapshots 4958–4960 determine usedFiles and deletion candidates during concurrent tasks. Done means historical data remains intact when expiration and orphan-file removal overlap.

Written by the indexing model from the issue text.