The Statistical Structure of Human Speech Sounds Predicts Musical Universals