Make diversity measures more efficient

There must be a lot of redundant calculation after computing the distance between sequence `A` and `B`, since the third sequence `C` must be almost identical to either one - we should be able to reuse a lot of the first calculation.