FEP on existing trajectory

The calculation of the Boltzmann factor with perturbation on the interaction potential should work in a rerun (compute fep).
The main idea of the FEP implementation is to do stepwise changes in the activation parameter (using fix adapt/fep), which I see as more suited to a calculation on the fly, since the perturbations should match. In a calculation using the FDTI method, then it makes sense to use reruns to try different deltas for the finite-difference.