Refactor optimiser interface to parameter sensitivity analysis

[BradyPlanden]

Currently the parameter sensitivity analysis is a method with the cost class and is called by the optimiser class during the _run() method. This allows the following:

• A user can add sensitivity analysis to an optimisation task by passing the `compute_sensitivities' argument to the optimiser on construction.
• The sensitivities will be printed to the user within the OptimisationResult output.

However, the interface to the optimiser via the args compute_sensitivities and n_sensitivity_samples is rather clumsy, and could be cleaned up for clearer usage. I suggest we do the following:

• Clean up the optimiser call compute_sensitivities to accept a tuple of format (Bool, Int) corresponding to whether the sensitivities are computed and the number of samples used. At the same time, depreciate the argument n_sensitivities_sample.
• Move the sensitivity calculation inside the OptimisationResult class to avoid mixing the calculation between the two.

Other thoughts are welcome, the main functionality I'd like to keep is the sensitivities stored and displayed by the OptimisationResult class.