v1.1.1

Changed the per read probabilities to float, to save space, as there shouldn't been that many alignments for one read to make a difference. Also used smaller int types in the read struct. This should be a large space saver for read classification when considering hundreds of millions of reads.