Text this: Optimal sparse coding of song in a size-constrained auditory system?