Add suggested torsions from Chris Bayly

[davidlmobley]

From Chris Bayly, here's an interesting torsion suggestion:

I have an OpenFF torsion dataset suggestion: s1ccc(Br)c1c2ncccc2 with s=o,n,s ; Br=H,C,F,Cl,Br ; n=c,n. This suggestion is based upon the 2015 Schrodinger FEP dataset for PTP1B where I think the interplay between the phenyl-thiophene torsion per se and the sterics vs electostatics (sigma-hole) non-bonds would be a good set of data for training all three.

It will have some similarity to @ChayaSt 's biphenyl set, but not the same. It's a really interesting interplay between sterics, electrostatics, and the sigma hole...