Slight rotation of molecule before computing Hessian

[TyBalduf]

This might not be a bug, but I couldn't find an explanation for why this was done. Before computing the numerical Hessian, several small rotations are applied to the molecule:

xtb/src/hessian.F90

Line 143 in 41059b4

call rotmol(mol%n,mol%xyz,step,2.*step,3.*step)

.

This leads to (small) deviations in comparison to manual calculation of the numeric Hessian or computing it via the C-API. Is there a reason this is done?