OVL updates

I mucked around with OVL a bit as part of my benchmarking runs to make the output easier to track. This also includes at least one bug fix. The changes are minor and should get into master asap, or as soon as my comparison between the old OVL and the new (bug-fixed) OVL is complete.

This fixes a bug introduced by the last change to OVL. We updated a shared reference because utils.remove_segments does not return a new object, so now we first make a copy of classifier_data.segs before training. I also updated the doc-string/comments for utils.remove_segments to hopefully avoid confusion about this moving forward.

I also added some logic to how Vetolist.reorder works so that smaller windows and larger thresholds are preferred in case there is a tie between metrics.

OVL and DOVL's print statements within _recalculate have also been updated to make them a bit easier to read.

Finally, I stripped out some unused functions from OVL and DOVL. They'd been deprecated a while back and the syntax within them was not really consistent with the rest of the objects (i.e.: with _recalculate). Instead of trying to maintain them, I decided to just get rid of them.