another layer off the onion: calculate_doublet_performance vectorized. code is...

another layer off the onion: calculate_doublet_performance vectorized. code is now a bit slower, but things should speed up once each scalar subfunction is vectorized as well