[cpp-threads] modes, pass 2

[Doug Lea]

 > Why don't you switch to hoist/sink stuff labels? Ok, perhaps with
 > "msync constraint(allow_reordering pre, allow_reordering post)"
 > calculator, so to speak. ;-)
 >
 >

My goal was to explain these in a way that minimized need to
model internal processor mechanics, mostly because they can vary.
This works out OK except for StoreLoad, which does require explanation
of the typical write-buffer mechanics; and also except for isync vs
LoadLoad on PowerPC, for which the fit is
inexact no matter how you try to explain it.

But you are right that someday I should redo this using
pre/post-reordering terminology. It would come out a little simpler.

-Doug