The quality of training datasets needs to be validated before training. Example: explosion of null values for a variable compared to a threshold or previous run, outliers detection, etc.

This task is to implement a set common tests that can be run on the output of training dataset pipelines, starting with the concrete use-case for the revert risk model.