Chad, this is an excellent reply. Thanks,
NativeClient is mostly about using the local processor to cut out server round-trips for computations. Unfortunately for your idea, it doesn't address the idea of the display engine; native client nexe's are expected to use the browser's ui. There is some experimental 3D support for nexe's via their alternative to NSAPI, but this is far enough away from Emacs that a port would be complex. For your goals, I'd guess that xembed (ala the xembed branch) is an easier practical path to a prototype.
This idea, like Tom Tromey's `rebase emacs on Common Lisp', Steve's `Emacs in JavaScript', and the various Guilemacs attempts point out (yet again) something I would not have guessed when I started with emacs ~22 years ago: while the most technically valuable piece of Emacs (aside from its freedom) is the extensive libraries of elisp code, the most critical piece is actually the display engine. It's very hard to move Emacs anywhere that the display engine won't go, and it seems to be very hard to move the display engine.
*Chad
P.S. This makes the herculean efforts of things like epoch, multi-tty, and bidi all the more impressive. Thanks again to all the wonderful emacs hackers out there.