Build olpc3-8 dies in early init.

[John Gilmore]

> 1) upstart wants to be pid 1. We reserve pid 1 for our own use. We can
> remove the conflict with a dirty hack, contained in the upstart RPMs at
> http://dev.laptop.org/~mstone/olpc3-repo.  Probably, we need to
> implement --init for upstart. (Or not use upstart.)

Since our PID 1 isn't really doing anything now, it might as well
exec upstart when it's done.  That would be a trivial change that
would get it running.  If and when somebody wanted to make the
sleep loop do something, it'd be THEIR job to mess with upstart.

(In addition, it would save about 5MB of RAM, which is currently
wasted on an unshared Python process that awakens every half hour
and then goes right back to sleep.)

And you'd shoot #4399, a serious performance bug that's been sitting
for 7 months, and #5809, 4-month-old way to panic the kernel trivially
from the keyboard.

	John