[flink] Add reliable cleanup result summaries to remove_orphan_files

### Search before asking

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

### Description

Users need to distinguish planned cleanup from actual deletion and determine whether a cleanup run covered all requested targets. Job completion alone does not provide this information, especially when metadata is unavailable or some targets cannot be scanned.

Separate planned and actual cleanup counters, account for incomplete scope coverage, and emit bounded, versioned result logs with a run ID and per-subtask summaries. Validate counter consistency and ensure dry-run results report no actual deletion.

### Willingness to contribute

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

Locate the remove_orphan_files entry point and trace how cleanup scope, deletion outcomes, dry-run handling, and existing logs are represented. Define tests for planned versus actual counters, incomplete coverage, bounded versioned summaries with run IDs and per-subtask data, counter consistency, and zero actual deletions in dry runs; done when those behaviors are validated.

Written by the indexing model from the issue text.