Allow for variable model types within the same Skeleton

[wtbarnes]

Currently, it is assumed throughout the codebase that every loop is defined using the same model and model interface. This assumption should be relaxed such that a single Skeleton can have a mix of models in a single skeleton. This could be useful, say, in using a simple model to create a background and then modeling a selection of loops a more sophisticated model.

One option would be to just create two (or more) distinct Skeletons, load their loop simulations, and then add their loops together to create a new Skeleton. However, there are still places where it is assumed that all loops in the Skeleton have the same model. I need to go through and double check and make sure I understand where all of those places are.

[wtbarnes]

Actually, it seems the only place this occurs is load_ionization_fractions which assumes that all loops have the same model interface anyway. So maybe there just needs to be a check there to ensure that all the model types are in fact the same

[wtbarnes]

Another consideration and probably the most difficult one to resolve, is the differences in simulation times between different models. As long as they are over the same interval, this is fine as they'll just be interpolated to the instrument time. But in the case of something like a background model where some of the loops don't evolve with time, we'll need to come up with some other way to deal with this.

[wtbarnes]

Maybe the easiest way to do this is to just define a compound model interface who's load_loop_simulation method just hands off the loading to some other set of pre-defined interfaces based on some properties of the loop (e.g. length). This could be illustrated in the example gallery.