Previously we would seek to the beginning of the io.ReadSeeker. It is much more
flexible to assume that the caller passes the ReadSeeker in an appropriate state.
1) correct a bug with the mdat atom.
Sometimes, the metadata can be at the end, after the audio
Demo :
get any music file and strip the metadata
convert it to m4a with ffmpeg
analyse it with MusicBrainz Picard
et voila, you have the metadata at the end.
Example of structure of a m4a created that way;
https://www.dropbox.com/s/602yivdstan7506/atom.txt?dl=0
2) add support for the "----" tag
See http://atomicparsley.sourceforge.net/mpeg-4files.html
It gives access to the tags created by MusicBrainz Picard
