Not very difficult to do now I've got my Pi brick sorted.
Yeah, now all you have to do is all of that pesky speaker and amplifier design
You've worked miracles with the Pi Brick, so I have faith in your ability to resolve tough issues, but the "analog" component is non-trivial, and there are some extra hurdles yet to cross.
MC for Linux has no ability to take an input stream at this point. So if you want to take advantage of JRiver's DSP for active correction (which is half the advantage of having a speaker embedded JRiver instance, in my opinion), the speakers would have to effectively be "wireless only." You could bypass the Pi entirely for wired audio input, but then you really have to throw a lot of engineering time at getting the speakers to sound great with no active correction, which is a non-trivial engineering challenge.
You'd also need to solve the sync issue between a pair of stereo speakers outside of the airplay context. You could always have a master-slave wired setup, e.g. one pi for two speakers, but wired to both of them (which is how some manufacturer's deal with it), but it would be a little odd to have wireless only speakers that need to be wired to each other
(unless you go for the boombox form factor, which is what I recently did for a pi-project in my daughter's room)
In all seriousness though, if you do wind up prototyping JRiver powered speakers, you should work on finding a way to take advantage of JRiver's DSP stack to do active correction. I'm consistently amazed at what I can do with cheap-o speaker components by throwing DSP at them (like I described over in my measurement guide). You can't fix timbre or the room, but you can fix or improve almost anything else. At this point the only reason that Pi's are proliferating around my house like dandelions is because I can make a tiny .06 cubic foot speaker with a $12 paper cone sound better than some $150 speakers I've heard.
Like I said, I have faith in you (and I'm sure you've built your share of speakers), but if you do go down this road and want a sounding board, I've built a lot of speakers (about a dozen from scratch, over two dozen counting major rebuilds) and would be happy to help out with a consult if I can.
I've actually started trying to rig up MC on a pi as a standalone DSP processor (analogous to a mini-DSP), as that market is currently saturated with overpriced tech that underperforms, but the main thing holding me back is that MC Linux can't take an input stream (or at least I haven't figured out how to rig it up).
MC for windows could do it, but embedded PC's aren't quite there yet (all the win 10 for IoT on the Pi2 fluff notwithstanding).