Don't save index if it's small enough

[staltz]

Forked from issue #5, from this comment.

The idea of saving how many results the index contains together with the index is a great idea! We already store the seq because we need to know when to update it, so it should be easy to add. Please note it is not count, its the size() function on the bitset. If you do a query on 3 parameters where you have an index of say the first, then you check how many results that have, and if its low enough you just filter directly instead of creating indexes for the other two as that is the slow part. What the threshold should be could probably be configured when loading jitdb, even better if it could be based on how many records can be read in 200ms or something like that.