This is branched off of #1328, for which the query language code was refactored and env_t was made reentrantly usable.

We could process more parts of a query concurrently. For example, mapping on the result of a range query, if it involves grabbing data from another node, could be done concurrently? How is other stuff like joins implemented? There's a chance that in the past 5 months some of this has been improved.