Support VectorPostprocessors for likelihood estimation

[ke7kto]

Reason

Many experiments are recorded in a manner that gives multiple simultaneous measurements, such as concentration profiles, line profiles, etc. These measurements are inherently independent, and it would make sense to use all of the available data in a stochastic model evaluation. A normal postprocessor value does not have the capacity to report all of the relevant information simultaneously.

Design

Accept VectorPostprocessor input for samplers/Likelihoods. Flatten the resulting std::vector<std::vector<Real>> and forward on to the rest of the decision-making process. Assume that experimental measurements corresponding to the VectorPostprocessor are listed in the document referred to by file_name in order. Throw an error if this is not mathematically possible. Alternately, require N columns, in the experimental measurements file, where N is the length of the VectorPostprocessor.

Impact

Greatly increase versatility of the stochastic tools module, though it may have wide-ranging impacts to implement.