Add function to handle side effects of fixation removal

[molpopgen]

Adding site tables (#211) creates side-effects for simulations where fixations are "pruned". The required operations should be abstracted out into a function in fwdpp::ts.

[molpopgen]

Note for future self: we need a function to prune fixations from the tree sequences. There are two strategies one can envision:

1. Remove mutations that lead to all alive nodes and zero preserved nodes. This method removes a subset of fixations.
2. Remove all mutations that are on the root nodes of marginal trees with only 1 root. Without ancient samples, this accomplishes exactly what we want w/o the need for sample list updating.

[molpopgen]

I want to delay this until after #225 and #283 are dealt with. This one requires thinking.