For the problems modelled with OptimalControl, return only the DOCP and then get the NLP with the associated method.

Add a description of what is the DOCP in the introduction of the documentation. It is somehow the OCP and the NLP.