v0.1.6

Bugs:

• Fixed numerical error occurring on high-trials multinomials, which caused the deletion of components

Features:

• Now you can use an outlier component for catching all the outliers, specify its estimated weight, and its hyper params (depends on data and prior, a NIW example will have very high pseudo counts and very wide isotopic variance).
• fit(...) now return the sub labels as well.
• cluster_statistics(points, labels, clusters) is a new function which, given the output of the model (or of other points/labels), provides the avg LL and probability of each cluster/