Implement gradients

[A-CGray]

To enable optimisation of FEMpy models, I need to implement gradient computation, using the direct or adjoint methods:

I therefore need to implement the concepts of functions and design variables along with the derivatives of functions w.r.t state variables and DV's and of the residuals w.r.t DV's. Most likely, geometric DV's will have to be handled separately from DV's associated with constitutive objects.

Once these partial terms are implemented, I will also need to implement methods for solving the adjoint or direct problems and computing the total derivatives.