Finding "next most close" MRCAs

[hyanwong]

It could be that the most recent common ancestor between a position at U and a match at V is only one of several possible common ancestral regions in V, the others of which are slightly more distant. For instance, there could have been a duplication just a little older than the coalescence point, forming 2 "similar" regions between U and V.

At the moment, when we use find_mrca_regions to look for similar regions to impose recombination breakpoints, we only find the most recent, but there is an argument that we should also be finding "reasonably recent" ancestors too, and matching against those, albeit with a lower probability.