Length of trajectories vs number of models in the shooting problem
Currently we have asserts (e.g. https://gepgitlab.laas.fr/loco-3d/crocoddyl/blob/devel/crocoddyl/ddp.py#L61) which force the length of the trajectory to be equal to the number of models in the shooting problem.
This is not true if we add impact models at the end of a contact phase. I'm currently working to remove these asserts and solve the other errors associated with len(models) > len(xs)