Use subselect for row_to_json so we don't have to rename keys

As described in http://hashrocket.com/blog/posts/faster-json-generation-with-postgresql we can (and probably should) use subselects to generate responses via the row_to_json in the query-recs in the query-engine which have proper key name so that we don't need to rename keys in clojure code, e.g. `f1` -> `hash`.