Propose an outline of how a map/reduce command will be translated into an aggregation. Will it just directly call the aggregation command after constructing some BSON request? Or will it enter at a lower level? What things will it skip if it does enter at a lower level?

As part of this we should also consider how a distributed MapReduce will work. Should it enter through cluster_aggregate.cpp or utilize some of the same helpers in cluster_aggregation_planner.cpp and sharded_agg_helpers.cpp.