AndHashStage uses an unordered map to store the records from the first child. It tracks the memory this consumes and compares it against a limit defined in a variable inside the cpp file. We will convert this to a query knob and spill the data when the memory exceeds this limit instead of spilling.

Additionally the unordered set it uses to track records already seen should be replaced by a RecordIdDeduplicator and spilled if necessary in order for the memory consumption to drop below the memory usage limit..