numerically differentiating vector valued functions

[Akshay5312]

A way to numerically approximate the jacobian given a vector valued function, (or the gradient of an Rn -> R function) would be beneficial

[tomstewart89]

Just FYI, it should be possible to generalise this to differentiating functions of $R^n \rightarrow R^m$ such that we can twice differentiate functions to get a hessian. See here for an example (this doesn't exploit the symmetry of the hessian unfortunately but :man_shrugging: )

[tomstewart89]

Done with #58