Add Hamiltonian Monte Carlo to Flarestack

[sgriswol]

It appears that most of the code needed to perform a fit on source weights, spectral index and number of source weights is already included in the flarestack.core.minimisation.FitWeightMinimisationHandler class. The minimization is performed in the run_trial method, and is called during execution of submitter.analyse(). Another class that is relevant to this execution is the flarestack.core.multiprocess_wrapper.MultiProcessor class. Two methods of this class in particular are relevant to creation and execution of the minimizer, __init__ where process instances are created and started, and run_trial which ultimately calls run_trial and performs the minimization.

In short, adding something to the effect of the following should put us in the right direction towards adding HMC to Flarestack.

@MinimisationHandler.register_subclass("fit_weights_hmc")
class FitWeightHMCMinimisationHandler(FitWeightMinimisationHandler):
    def run_trial(self, full_dataset):
        ...
        return ...

[sgriswol]

A brief procedure for achieving this could be

1. Put HMC into run_trial function for the new class
2. Get the example notebooks running with the new minimizer
3. Run comparison of our minimizer to example minimizer (sanity check)