Integers are inserted into the index in a little endian order. We should use a big endian ordering so that we can support optimizing less and greater queries on numbers. This is most likely a file format breaking change that requires rewriting the indexes, so the work should go to next-major. 

This is part of phase 1 geospatial because we want to make sure that we have prepared the way for using the index for geospatial values later on (possibly in phase 2).