This would be very simple for soundcloud to do. Also, I don't bother listening to tracks with soundcloud because its autoplay feature isn't able to understand that if I'm listening to instrumental piano versions of video game music, then please don't blast me with some super loud chiptune music next.
As a SoundCloud user, I would argue that a great number of the target audience (musicians) would be turned off by having a carefully crafted mix altered. Anyone who has spent large amounts of money on the tools of their trade has an investment in achieving a particular result from their hard work. The listener is (IMO) secondary to the artist in this case.
Good luck with that.
I would revert to hosting my music only on my own sites if I learned that SC was altering my audio.
Edit: imagine if github altered code when hosting public repositories.
If you're on a windows box, most audio drivers (at least in the last few years) have some form of system-wide volume normalization. Of courses it's hidden in some oddly named tab under the audio device properties.
I used to just run one on my computer, but it usually involves installing a pretty hacky/custom audio takeover program and isn't really accessible for most people.
Purists will probably complain about the audio being modified. I suppose you could either (optionally) enable it dynamically, or you could store separate tracks for the unchanged and changed audio. But I suspect that most people would prefer their audio to be normalized. I like to throw in a good multiband compressor when I have one too, but that may be a little too opinionated for the general use case.