Fix rectangularizing featureset w/ unsorted labels

[bnaul]

Small (bug?)fix: the main work of rectangularize_featureset is done by the Pandas unstack function, which automatically sorts the dataframe (by row); this leads to unexpected behavior when e.g. checking model predictions against a list of known targets. This change explicitly undoes any sorting and preserves the original ordering of featureset.name, which I think is how one would expect it to work.

[stefanv]

Thanks for clarifying. :+1:

[acrellin]

:+1: