Cell optimisation without neutral atom potential can fail

[davidbowler]

The Ewald routines which are only called when the neutral atom potential is not used are not updated when the cell size is changed, which leads to incorrect results and can lead to crashes. We need a new routine to recalculate the various contributions.