arm: don't use atomic operations to gate non-boot CPUs

arm: don't use atomic operations to gate non-boot CPUs

Since the cache is not enabled that early, better not to rely on
load-linked/store-conditional.  Instead, have the boot CPU call the
other CPUs by their IDs, just like we do later for proper CPU bringup.

Signed-off-by: Tim Deegan <tim@xen.org>
Committed-by: Ian Campbell <ian.campbell@citrix.com>