Problem

When closing async record cursors (e.g., AsyncWindowJoinRecordCursor, AsyncFilteredRecordCursor, AsyncGroupByRecordCursor), the current implementation reduces/executes all published tasks before completing the cleanup. This is wasteful because these task results will be discarded anyway.

Current behavior

• On close(), the cursor calls collectCursor(true) and frameSequence.await()
• This causes remaining tasks in the frame sequence to be reduced (executed) by workers
• The results are then discarded since the cursor is closing

Suggested enhancement

When closing a cursor, skip all tasks that belong to the frame sequence being closed instead of reducing them. This would:

• Avoid wasting CPU cycles on work that will be discarded
• Reduce query latency, especially for queries with LIMIT n that close early

Additional context

The SIGSEGV bug in #6662 manifested reliably with LIMIT n clauses because the query owner thread (when not a worker thread) was guaranteed to reduce remaining tasks during close, which accelerated the race condition between task reduction and resource cleanup.