The vorbis decoder doesn't handle more than one audio frame packed into
the same AVPacket, so they need to be split in the depacketizer.

Originally committed as revision 24704 to svn://svn.ffmpeg.org/ffmpeg/trunk