Revisiting Acoustic Similarity in Emotional Speech and Music via Self-Supervised Representations