Internal NJ tree moves

[ms609]

cf https://www.biorxiv.org/content/10.1101/2024.10.28.620561v1.full.pdf

Machine learning is probably redundant for phylogenetic datasets.

Could do even better than Neighbour Joining with biorxiv.org/content/10.1101/2023.10.11.561902v2

Would probably need to be part of a broader rewriting of tree structure and thus SPR/TBR moves – which might also consider restructuring TBR as a pair of SPR moves.