mutagen.id3 needs:
 * Detect tags in non-beginning locations, particularly if multiple
   separate version tags exist at the beginning of the file
 * Perhaps implement tag merging - "intelligently" merge information
   from all known tag formats
 * Some test cleanup

mutagen.apev2 needs:
 * General cleanup to fit the basic structure of the rest of Mutagen.

Profile Hotspots:
 * __determine_bpi takes 50% of TMP3,TEasyID3; BitPaddedInt takes 25%.
 * OggPage.replace takes 40% of Ogg tests; OggPage.write takes 10%;
   the rest is from our mmap fallback.
