Add jacobian to model optimisation
We currently rely on numerical estimation of the Jacobian when optimising models. However, an analytic form for the Jacobian can be determined (at least for the (extended-)Tofts Model and DIBEM - 2CXM may be more complicated).
We should make model classes able to define a Jacobian, and provide the option of using this during optimisation. This can tie in with issue #25 (closed) - providing options for choosing optimisation.