Generalizing 'collapse_mutationless_edges' and small fitchcount bugfix

[richardyz98]

This pull request includes the generalization of the collapse_mutationless_edges function. This allows us to use a function that does not factor in missing data when determining whether to collapse two nodes. If we do not collapse on missing data, then the baseline score increases for trees as missing data increases.

Additionally, it fixes a small bug in fitchcount which can error if no root is provided, such as on an initial call of the function.