Improvement suggestion: res-vs-fitted plot

[lauravana]

We can consider introducing functionality for plotting the different residual types (pearson, deviance, etc) for GLMs for the case of non-Gaussian families.