Zephyr is migrating into hardware where memory ordering is starting to
matter. The existing gcc-based defaults have always been specifying
__ATOMIC_SEQ_CST, which is safe and correct. And all the
arch-specific assembly we have currently are either safe or for
platforms where barriers aren't needed.
Discussion in #42831 made the case that it would be nice if we were to
formally promise this, and require it from future implementations. So
let's do that.
Fixes: #42831
Signed-off-by: Andy Ross <andrew.j.ross@intel.com>
24d4becac2