Further validation in check-model?

[lumip]

twinify-tools check-model currently only verifies that models work with twinify, i.e., that there are no errors in actually running the program. We could add some additional validation of the model structure/inference itself (at least verify that inference is able to capture the prior or something like that?)