Implement "backrub" algorithm for rotamer optimization

[tic20@…]

A less pressing potential enhancement would be to include the "backrub" algorithm when optimising the rotamer for clashes/fit to density (​https://academic.oup.com/bioinformatics/article/24/13/i196/233217). It's quite simple to implement and often makes a big difference, although it still doesn't help in every case (in particular, a 180 degree flip around chi1 of Ile, where you really need a degree of freedom waggling CB ~parallel to the N-C axis to avoid the sidechain staying obstinately out of density). I have a basic version in ISOLDE, although I need to do some more intelligent weighting of density vs. clashes and prior probability before it'll be really useful.