Performance improvement

[ClementBeal]

The library is pretty quick but it's still very slow for users fetching metadata from huge library (~40 000 tracks). I don't really know how to improve significantly the performance.

Here is a list of things to check. They might improve the performance :

• Make a more efficient Buffer class
• The parser selector could be based on the file extension. If it fails, the callback would be to get the first 40 bytes and match a magic word
• Replace sublist() with .view(). The library is returning a lot of UInt8List() that are extracted from another Uint8List instance.
• ID3v2 parser has an unefficient way of finding the Xing header.
• ID3v2 String.fromCharCodes() is called a lot. Perhaps we shouldn't translate the bytes into String and compare array of bytes instead of string.