Consider switching between merging and iterating based on query size

[tomwilkie]

See #713 (review)

[gouthamve]

We've looked at this in our bug scrub and decided to close it because of the complexity it adds to the code. The speed up is for smaller datasets which are fast enough anyways. We will revisit this if required.