The difference between MIDI and Audio-Files (e.g. WAV, MP3, ...) is that in an audio file, there is saved the way the speaker's membran has to move. To describe this as exact as possible, it needs normally 44100 samples at every second. That's very much => the file is quite big.
MIDI just saves musical notes, it says something like "play tone c over a duration of a second using sound grand piano". This method only requires much less samples per second, so the file gets smaller. Of course it needs a synthesizer to be decoded and it always sounds different depending on the synthesizer used to play it.
Since Warcraft does not include a synthesizer, it needs audio files for sounds. Just as explanation for you.
