Add more goodness of fit functions?

[drbenvincent]

The stats submodule has some useful goodness of fit functions, like r2_score. I wonder what the appetite would be to expand this to include metrics such as:

• mae
• mape
• maape
• ecdf_crps - Kruger, Lercha, Thorarinsdottir, Gneiting (2020) "Predictive Inference Based on Markov Chain Monte Carlo Output" https://arxiv.org/pdf/1608.06802.pdf Eq. 9 Page 11.
• mdape

Tagging @lucianopaz because of past interest in implementing some of these.

[aloctavodia]

It makes sense to add those. I recommend adding them directly to https://github.com/arviz-devs/arviz-stats