Current implementation of FlattenRecordBatch does not create transfer pairs with the target vector named with the desired field reference. This seems working fine with repeated types as the upstream project or ArrayPathSegment makes the renaming implicitly. Overall relying on upstream makes the code fragile. Also,
DRILL-2150 requires this patch as it alters the way repeated vectors are modeled. This patch proposes to implement a less fragile transfer pair handling via explicit naming.