Transforming samples requires application to dicts

[odunbar]

Issue

We should add the functionality to transform parameter dicts, not just arrays. This should be straightforward as we must use the prior to perform the transform, and the prior also contains the parameter names.

Example setting:
Currently, to transform samples from a Samples distribution, we must convert it into an array first with the correct ordering

# S, Samples distribution (e.g. a posterior), 
# P, distribution (e.g. a prior with same parameter names)
samples_dict = get_distribution(S)
param_names = get_name(S)
samples_arr = zeros(sum([size(samples_dict[n],1) for n in param_names]), size(samples_dict[param_names[1]],2))
for (i,k) in enumerate(param_names)
        samples_arr[i,:] = samples_dict[k]
end
constrained_samples = transform_unconstrained_to_constrained(priors, samples_arr)

Suggested solution

This is what we would want to transform samples from a Samples distribution

# S, Samples distribution (e.g. a posterior), 
# P, distribution (e.g. a prior with same parameter names)
samples_dict = get_distribution(S) #a Dict(param_name => samples_for_name)
constrained_samples_dict = transform_unconstrained_to_constrained(samples_dict)

Which can be done by adding the code snippet into functions transform_constrained_to_unconstrained(pd::ParameterDistribution, d::Dict), transform_unconstrained_to_constrained(pd::ParameterDistribution, d::Dict) into ParameterDistributions.jl