Check MCMC convergence

[dp-rice]

Charlie suggested that we report R-hat to assess the convergence and mixing of our posterior samples. Here's a snippet of code that calculates it for our parameter b, the overall P2RA coefficient (not the location-specific coefficients):

import arviz as az

model.fit_model(num_samples=num_samples)

coeffs = model.get_coefficients()
rhat = az.rhat(
    coeffs[coeffs["location"] == "Overall"]
    .to_numpy()
    .reshape((num_samples, -1))
    .T
)

@simonleandergrimm , you'll have to add arviz to the requirements, add this to fit.py, and then decide how/where you want to report rhat.