`average_odds_difference` metric is wrongly represented as an equalized odds relaxation

[AndreFCruz]

The average-odds metric defined here is listed as "A value of 0 indicates equality of odds."

This doesn't seem to be true, as any pair of points on the following line would have average_odds_difference=0 but not fulfill equalized odds necessarily:

This mistake seems to be repeated on the IBM website here