There is now a helper function to run bulkWrite request under an internal transaction. The bulkWrite command returns a cursor in its response. For the case where the response cursor isn't immediately exhausted (i.e. cursor id != 0), we must add logic to iterate through the response cursor in the transaction api helper.

This only helps bulkWrite used in internal transactions. I think it is ok to punt on this until we have real use cases to run big enough bulkWrite in internal transactions that actually requires multiple response batches.