Simplify tree sequence to remove invisible recombinations

[hyanwong]

It might be useful to have a routine that can be run on a TS that looks at the distribution of mutations and removes nodes (recombinations) that are "invisible", given the set of variants.

[jeromekelleher]

Another way of looking at this is to minimise the tree sequence: how many nodes/edges can we throw away while not losing any essential topological information?

[jeromekelleher]

See #545 for initial version.

[jeromekelleher]

I think the reduce_site_topology feature in #545 does solve this fully, so closing.