Rename rater_agreement_pearson() to rater_agreement()

[hagenw]

Closes #15

Renames audpsychometric.rater_agreement_pearson() to audpsychometric.rater_agreement(), to have the option to use another correlation measure at a later stage, e.g. by adding an association_type argument.

[ChristianGeng]

I am starting to think that there is even more generality to this:

• the association measure is one thing, and by providing a more general name suggests that we could increase the flexibility on that end.

• However, there is even more: there is the variable average_ratings_other_raters that is hardwired to use the mean value. The mean is a kind of "gold standard" in the old terminology. But could it not be that we want the mode, or even the ewe?

Apart from that: using standard association measures like pearson runs against the spirit of reliability coefficients that account for something like guessing behavior on the annotator side in the case of non-numerical data. For numerical data there is the icc family of coefficient that deal with these problems - by using aov decompositions that decompose variance into its components (between items, between raters), with different decompositions answering different questions.

Possibly we should deal with these generalizations later?
For now I would approve.