update mapping algorithm

[ekg]

A big performance win lies at the other end of a small optimization. If we can guess what the best mapping target is on the basis of kmer matches, we can try to run fewer gssw alignments. If this can happen we can run a lot quicker when mapping.

One idea would be to measure the informativeness of each kmer, and then build a conditional entropy metric for each mapping target. We want to evaluate the mapping targets where our kmers hits are rare, and where there are many hits, before mapping against a huge complex of lower quality hits.

[adamnovak]

We've improved the mapping algorithm significantly since this was opened, and we're in the process of improving it even more. So this is complete.