Sklearn bugfix

This fixes some bugs introduced by switching to ClassifierModel within the sklearn stuff, mostly associated with accessing some attributes of the model object.

I've also changed OVL's behavior so that it will read in and cache all relevant triggers before it begins iterating over channels. This will prevent repeated reads of the same multi-channel file and has been shown to greatly speed up the I/O needed for training. In addition to this, I've added a new verbosity flag to make OVL print out the progress of its caching.