Reversion push code needs to deal with recombinants

[jeromekelleher]

One of the largest artefactual recombinants in the alpha version of the sc2ts trees (node 601947 in upgma-full-md-30-mm-3-2021-06-30-il.ts.tsz) is the direct result (seemingly) of reversion pushing not dealing correctly with recombinants.

[jeromekelleher]

In principle there's no problem with the current approach being applied to nodes which have two or more parents, it's just a little more involved and I skipped it in the initial version.

[jeromekelleher]

This looks like the same issue here (4 mutations, all immediate reversions near a recomb node)