A first iteration for quality estimators for signal noise

[stasinos]

Implement the APIs from Issues #2 and #3 for the following use case: inter-annotator agreement is used to recognize noisy signal. Then train a model that recognizes noisy signal.

[stasinos]

Folded on category "ST", "WW", "AP" and trained a decision tree. Accuracy is not bad for a first shot:

ST -> WW, 7658 / 11730
ST -> AP, 6175 / 7886
ST -> ST, 21790 / 30430
WW -> ST, 21274 / 30430
WW -> AP, 6181 / 7886
WW -> WW, 7882 / 11730
AP -> ST, 21359 / 30430
AP -> AP, 6188 / 7886
AP -> WW, 7796 / 11730