Consolidate BayesianProblem, Posterior, JointDistribution etc.

[nabriis]

Currently BayesianProblem is responsible for methods that act on classes like Posterior, JointDistribution etc.

I may be worthwhile to reconsider this approach and put the methods directly into Posterior, JointDistribution etc. to reduce redundancy.

• JointDistribution: Sample (forward sampling)
• JointPosterior?: Sample (MCMC sampling)
• Posterior: Sample (MCMC sampling)
• Distribution: Sample (Direct? Maybe not always possible)..

Further considerations: Should we distinguish direct and MCMC sampling?

• [ ] Design
• [ ] Implementation

[amal-ghamdi]

should user only use BP and extract JointDistribution from the BP