The new simplified __pvclock_read_cycles does the same computation
as vread_pvclock, except that (because it takes the pvclock_vcpu_time_info
pointer) it has to be moved inside the loop.  Since the loop is expected to
never roll, this makes no difference.
Acked-by: NAndy Lutomirski <luto@kernel.org>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>