To everyone saying that changing the pitch and the speed simultaneously would sound horrible because it doesn't work with a DS, please educate yourself as to how sound works.
If you want a short handwavy explanation, read Halogen's post. This is a bit more technical if you don't believe him.
In short, all audio is just a waveform, it's a position on a single axis as a function of time. A graph where time is the X axis. But sound is not a function you can represent with an equation, so the computer instead simply takes points along every 'n' interval on the time axis, commonly 1/44100ths of a second. These points are then stored in a file known as an mp3 or a wav.
In order to reproduce sound, the computer takes the data points from the file, and plots them back on a new graph. It then scrolls through the points at a speed of 44100 points per second, which then fully reproduces the original sound.
The current granular algorithms to slow down the music is a problem because it results in major distortion of tone and loss of transients, which is a direct result of trying to process the sound in a way that keeps the pitch the same, which is a lot like trying to force a circular rod into a triangular hole. The largest problem with granular stretching is that the transients are lost since matching these transients is exactly what we are trying to do as charters.
What are transients you ask?
TRANSIENTS ARE THE REASON WHY RHYTHMS EXIST. If you distort or lose the transients, you distort or lose the rhythm. Simple as that.
The granular algorithm abraker used in his demonstration is a much better algorithm that what the current editor has. Unfortunately, it's very computationally taxing to do in real time and it still greatly distorts the tone and transients.
The OP's suggestion here is instead of using a badly written algorithm designed to transform and distort the sound data, we simply play the original data back at a lower rate than 44100.
So for 50% speed, we play the data back at 22050 points per second.
This will result in TWO THINGS:
1) The music will be half as fast.
2) The pitch will automatically be taken an octave lower. This is due to the fact that frequencies and tones are nothing more than repeated patterns of points in the data, and slowing the data down will also slow down the patterns, causing the halving of the frequency and thus lowering the pitch of everything in the music by an octave.
But most importantly, the transients will be preserved. This is why we want to slow down the music to begin with: to hear the transients better.
This is why lolcube's argument is painful for me to read. It demonstrates an inherent ignorance of how sound works.
Thanks for reading, hopefully at least people have some idea as to what OP is trying to suggest now.
tl;dr lolcube is wrong, please understand what OP is trying to say and realize it's a great idea which should be implemented.