Guaranteed latent model functionality

[SamuelBrand1]

          The implicit promise of this being a latent model is that it will work with the library of latent model functionality. Does this implementation do so? If no do we need to rethink the idea of making this a latent model

Originally posted by @seabbs in #510 (comment)

As @seabbs mentions above the latent model type (setting ODE model and parameter priors) offered by #510 might not make sense with the whole latent model stack. E.g. What does broadcasting a sample from this kind of latent model mean?

This opens up a conversation about what we want for the fundamental capabilities of latent model.

[seabbs]

Yes agree. As discussede f2f we likely need some kind of latent model type tree that dispatches between vectors, matrices and other

[SamuelBrand1]

Agreed: latent random variable (processes) we expect to be indexable should be one branch and others on the other.