Core: Don't fail when bulk deleting metadata in CatalogUtil

[sumtiogo]

I recently opened Issue #15456 regarding this missing exception handling. Since the fix is relatively straightforward and I have already verified it with a unit test locally, I went ahead and created this PR to facilitate the discussion and review process.

Fixes #15456

Why are the changes needed?

In CatalogUtil.deleteRemovedMetadataFiles, there is an asymmetry in how exceptions are handled between bulk operations and single-file deletions.

When the FileIO supports bulk operations, deleteFiles is called directly without failure suppression. However, implementations of SupportsBulkOperations.deleteFiles can and do throw exceptions (e.g., BulkDeletionFailureException in S3FileIO when S3 DeleteObjects fails). Because this cleanup operation is intended to be best-effort, a system-level or network exception here should be caught and logged rather than bubbling up and potentially crashing the main calling process. The single-file deletion (else branch) already handles this correctly using Tasks.suppressFailureWhenFinished().

What changes were proposed in this pull request?

• Wrapped the SupportsBulkOperations.deleteFiles call in a try-catch block within CatalogUtil.deleteRemovedMetadataFiles.
• Added a warning log to record the failure and the number of files that failed to be deleted, maintaining consistency with how exceptions are handled in other bulk deletion cleanups (e.g., BaseTransaction).

How was this patch tested?

• Added a new unit test testDeleteRemovedMetadataFilesBulkDeletionFailure in TestCatalogUtil using Mockito to simulate a RuntimeException during bulk deletion and verified that it is safely caught without throwing.
• Ran ./gradlew spotlessApply and ./gradlew :iceberg-core:test locally.