create pytensor RVs from missing datasets

[augeorge]

for a given model variable and condition, the data could be: observed (normal distribution RV), unobserved (laplace distribution RV), or excluded from the model (zero). we need a function to help with this