Constraints for multivariate prior in Bayesian update scheme

@ricardoV94

Is it possible to define the multivariate distribution purely numerically, by rejecting points that do not satisfy constraints? Or would that destroy performance beyond usability?