[Improvement] Reuse primary-key sorted index after partial source retirement

[wangyong9999]

Search before asking

• I searched in the issues and found nothing similar.

Paimon-cpp version

Apache Paimon C++ main at 4027601.

Minimal reproduce step

1. Create a primary-key table and compact data files A and B into a positive level.
2. Build a source-backed scalar index payload whose immutable source metadata is [A, B].
3. Commit an update which retires A, keeps B active with a deletion vector, and adds an uncovered replacement file C. The original payload remains active.
4. Scan an indexed predicate.

What does not meet expectations?

The reader currently validates a payload only when its complete source list equals the current active source set of the whole level. It therefore rejects [A, B] against active [B, C] and turns both surviving indexed data and the new file into ordinary scans.

The payload still has a valid immutable ordinal namespace for B. The reader should reuse it for the active intersection, preserve the deletion vector on B, and leave only uncovered C on the ordinary scan path. Multiple retained payload groups should be accepted only when their active source intersections are disjoint; ambiguous overlap must continue to fall back.

Anything else?

The fix can remain read-side only: no public API, storage format, or index protocol change is required.

Are you willing to submit a PR?

• I am willing to submit a PR.