Store old model state information

[antoniaih]

Unused model components are deleted in the model state. For example if the model has a CLMM parameter and the user set a covariate on CLMM then this covariate is deleted if the user changes the model to FO elimination, because CLMM does not exist anymore. When then going back to Michaelis Menten elimination it would be nice to be able to retrieve the information that there was a covariate on CLMM.

Maybe the model_state object could store unused information in a new attribute. That attribute could be a dictionary containing all model_state attributes.