vp9: Return the correct size when decoding a superframe