Jukes cantor likelihood

[willdumm]

This PR provides an implementation of Jukes Cantor likelihood on histories, with branch lengths inferred to match the expected number of mutations on each branch to the observed number of mutations on that branch.

Also included:

• implements __contains__ method so that history in dag returns whether history is among the histories in dag
• provides a way to estimate the number of bifurcating histories in the DAG. This was intended to provide a new way of computing node support by summing over all bifurcating trees, but those plans have been scrapped.

[matsen]

but those plans have been scrapped

Can you remind me-- this is because it just didn't work well?

[willdumm]

For now I've abandoned the idea because there doesn't seem to be a good way of counting bifurcating resolutions correctly, since the new nodes that would result from resolving a multifurcating node may already be in the DAG. I think there's something more involved that could be done that would involve changing the data structure slightly...