outlier_scores.py need fixing

In GitLab by @cavart on Jan 6, 2021, 11:04

It seems that:

a) find_prop_markers

c) get_confusion_zones_percentiles

b) get_confusion_zones_std

are not doing exactly what they are intended. See `ca/913-outlier_scores_functions.ipynb` in the notebooks repository for a demo and with basic critical thinking, the plots don't represent what the careless author says it does :)