[Fix] check_outliers() expose 'xi' parameter for OPTICS

[DominiqueMakowski]

I have a nice new dataset of a questionnaire validation that included attention checks. I have been playing with check_outliers(), and the OPTICS method. Interestingly, the current rule for its threshold that we have didn't work at all, + I had to change xi to actually make it work (hence this PR that allows to do that).

Interestingly though, we typically consider outliers observations with a large distance, but my data shows that people that have failed attention checks (i.e., "objective" outliers) tend to have a low distance:

(the logistic model failed ~ distance_optics makes it a significant predictor of failure)

I wouldn't recommend any action for now, maybe we can simply mention this observation (that it is possible that low distances are also problematic) once my paper is published and link to it

[strengejacke]

Looks good, can be merged, or not?

[DominiqueMakowski]

Yup, it should work! Thanks