Planned feature: codon sequence evolution models

[jeromekelleher]

We should be able to implement models of (independent) codon sequence evolution following the approaches of Arenas and Posada 2007

It's not entirely satisfactory that we'll end up with independent trinucleotides (i.e, each "site" has three bases) but I think this is fine for a first pass, and will help us to start thinking more clearly about the consequences of overlapping sites upstream in tskit.