The routine that did interpolation/extrapolation for the KS potential

during the time-propagation was horribly written (my fault), which did
not matter when propagating forward, but was introducing an error when
propagating backwards (for QOCT propagations). I re-coded it.