Grid update Bugs when trying to calculate box numerical derivatives

[carajillu]

Grid update works OK when calculating numerical derivatives of atom coordinates, but when it starts calculating the derivatives of the box, the distances gridpoint-atom increase unrealistically (a grid point that is at r=0.0742135 from each of atoms 1 and 2 suddenly appears to be at r=1.91191). This results in the bsite_bin vector being filled with zeros and the grid point position being updated to -nan, and then all the numerical derivatives are gibberish.

Not a problem if it just bothers when debugging derivatives, but need to double check that it does not mess with biased simulations (cross that bridge when we reach it). For now, found a workaround by not updating grid if numstep is larger than 3 x number of atoms