Likelihood calculation improvements: Add in model_responses_covmat for observed_responses_covmat

[AdityaSavara]

Right now, in creation of the likelhood

observed_responses_covmat

which comes only from the observed uncertainties, and is used to make simulated_responses_covmat

However, we also have the possibility of a separate model induced uncertainty. So we need to make another variable called

modeled_responses_covmat

Or something like that.

We will (for now) assume the uncertainties are orthogonal between observed_responses_covmat and modeled_responses_covmat

Since the user puts in "responses['responses_observed_uncertainties']" , perhaps we should make the analagous variable model['responses_modeled_uncertainties']. The code can check if it's an array or a list, and if it is neither, then can assume it is a function that will return the uncertainties as a function input parameter set and/or independent variables.

As a related issue, as of April 20th 2020, Still need to make responses['independent_variables'] along with non-linear regression feature to better create simulated_responses_covmat from observed_responses_covmat

[AdityaSavara]

During this process, replaced the old simulated_responses_covmat with comprehensive_responses_covmat. Then will use model['responses_simulation_uncertainties'] rather than model['responses_modeled_uncertainties'].

Should consider changing to responses['responses_observation_uncertainties'] as well.

[AdityaSavara]

I have added the feature in. But it's not yet compatible with the feature model['reducedParameterSpace'] and probably also not with responses['reducedResponseSpace']

So I have added this line comment in the code.
def get_responses_simulation_uncertainties(self, discreteParameterVector): #FIXME: Make sure this works with responses['reducedResponseSpace'] and model['reducedParameterSpace']. I don't think it does.