Generic likelihood specification using observed and expected values rather than Data, Prior, Random effects setup

[Andrea-Havron-NOAA]

To Dos:

• Create generic VariableObject to replace DataObject
• Add derived_quantity as option to estimation_type?
• Reconcile the difference between the backend VariableObject and front end VariableVector (dimensions, initial, final values, etc.)
• Collapse distribution type in the distribution functors and collapse code needed to differentiate between data, prior, and random effect observed and expected values (keep distribution_type though for ordering in model.hpp)
  • Collapse SetupData, SetupPrior, SetupRandomEffects calls in information.hpp
  • Collapse pointers (prior, re, observed_data, etc.) to a single observed and expected pointer
  • Simplify get_expected() and get_observed() functions
  • Collapse distribution function calls so there is not a separate loops for each type

Wish List:

• Collapse distribution functors to single child class with case/switch format