Create a function to fit a set of posterior samples into a given prior.

[aloctavodia]

We already have part of this functionality in PreliZ, but we could expose it to the user and make it nicer for users

The steps should be:

1. Read a model, and extract variable names and their families
2. Fit the posterior samples into the corresponding families (use MLE).
3. Generate samples from prior predictive distribution of the new model.
4. Compare the posterior samples to the samples from the new model

Optionally, in point 2 we could return the original family + a new set of families that better match the posterior samples.
For point 4 we could have a function to compute a set of visualizations like ECDFs. We could also compute a numerical metric like the Wasserstein distance...

[aloctavodia]

The functionality described in the first 3 steps has been added #508 and then extended here #510. We are still missing point 4. We will also need an example explaining this function, and very importantly the limitations of this approach