Adaptive time-step for surface hopping

[danielhollas]

SH routines needs to be revamped. After that one could implement adaptive time-step, similar in spirit to FMS approach. This should allow for larger time-step while also allowing for better energy conservation.

This could be well tested with new TeraChem-FMS MPI interface, which needs to be updated.
In MOLPRO, one needs to add diabatization to r.molpro to allow for monitoring of CI vectors along the trajectories.