Find mean of a posterior sample

[ms609]

"Centroid" algorithm uses tree search to find tree that minimises sum of squares to all trees in a posterior distribution
https://doi.org/10.1093/bioinformatics/btae514

Could this be (efficiently!) accomplished with arbitrary distances in TreeSearch? Could we explore variance etc as other statistical properties?

Probably the efficient answer from a CID perspective involves getting the whole process into C++ such that split similarities are cached and only recalculated when a TBR/SPR move changes split membership.