[core] [flink] [spark] Preserve blob view references during forwarding

[leaves12138]

Purpose

Blob view currently resolves BlobViewStruct to upstream BLOB content on reads. This makes it hard to forward one blob view table into another blob view table while preserving the original upstream reference.

This PR adds a dynamic read option to preserve blob view references during such forwarding:

• blob-view.resolve.enabled=true keeps the existing default behavior and resolves blob view fields at read time.
• blob-view.resolve.enabled=false skips read-time resolution and returns the original BlobViewStruct as a BlobView, so writing it into another blob-view-field stores the same upstream reference.

Changes

• Add CoreOptions.BLOB_VIEW_RESOLVE_ENABLED.
• Make DataEvolutionTableRead skip blob view resolving when this option is disabled.
• Add a test that forwards T1 blob view rows into T2 while preserving references to T0.
• Document the option and the forwarding pattern.

Tests

• mvn -pl paimon-format -am -DskipTests package
• mvn -pl paimon-api,paimon-common,paimon-format -am -DskipTests install
• mvn -pl paimon-core -Pfast-build -Dtest=BlobTableTest#testForwardBlobViewReference test
• mvn -pl paimon-core -Pfast-build -DskipTests compile
• mvn -pl paimon-api,paimon-core spotless:check -DskipTests
• git diff --check

[leaves12138]

Thanks for the update. I reviewed the latest patch and the blob-view forwarding path looks good to me.

The new blob-view.resolve.enabled=false read option cleanly bypasses the blob-view resolving reader, so unresolved BlobViewStruct values can be forwarded without triggering upstream lookup or materializing the actual blob bytes. The Flink and Spark row wrappers also preserve unresolved BlobView values via Blob.serializeBlob, while keeping the existing resolved/blob-as-descriptor behavior unchanged.

The added core/Flink/Spark coverage exercises the important end-to-end forwarding case, and the current CI is green. I do not see remaining blocking issues.

[JingsongLi]

+1