Update Pass / Fail Labels

[firste]

When discussing this package with health systems, I have heard feedback that given the upcoming OCR 1157 changes, the pass / fail language is too strong and makes them feel like they need to report to regulators when they see fails with their models. This coupled with the fact that models may have the specific attributes they want to segment by as input features means that fails are missing the correct context.

It would be great if this language could be configured on a per attribute basis so users have the right context when reviewing the results.