Generalize to multi-allelic mutations

[fedarko]

Changing this for just a single analysis is relatively easy; changing it for every step of the pipeline, and testing it at every step of the pipeline, will be much more involved.

Long story short, the ideas of a p-mutation or an r-mutation are currently defined in a binary way -- a position either is a mutation (aka "is mutated") or it isn't. But positions can have multiple alternate nucleotides, and ideally we would account for these rather than ignoring them.

It's simple to extend the idea of a p-mutation or an r-mutation to multi-allelic positions (so that a position can have 0, 1, 2, or 3 mutations, rather than 0 or 1 mutations). However, the effects of this change will impact:

• Naive mutation calling
• Computing & interpreting diversity indices
• FDR estimation and fixing
• Hotspot identification (I think coldspot gaps are ok as is, though)
• Read smoothing (I think link graphs are fine as is)
• Mutation matrices (account for multiple alternate codons)