runtime: improve CALLFN macro for ppc64x

The previous CALLFN macro was copying a single byte at a
time which is extremely inefficient on ppc64x. This changes
the macro so it copies 8 bytes at a time.

benchmark in reflect:
name                      old time/op    new time/op    delta
Call-8                       177ns ± 0%     165ns ± 0%    -6.78%  (p=1.000 n=1+1)
CallArgCopy/size=128-8       194ns ± 0%     140ns ± 0%   -27.84%  (p=1.000 n=1+1)
CallArgCopy/size=256-8       253ns ± 0%     159ns ± 0%   -37.15%  (p=1.000 n=1+1)
CallArgCopy/size=1024-8      612ns ± 0%     222ns ± 0%   -63.73%  (p=1.000 n=1+1)
CallArgCopy/size=4096-8     2.14µs ± 0%    0.53µs ± 0%   -75.01%  (p=1.000 n=1+1)
CallArgCopy/size=65536-8    33.0µs ± 0%     7.3µs ± 0%   -77.72%  (p=1.000 n=1+1)

Change-Id: I71f6ee788264e61bb072264d21b77b83592c9dca
Reviewed-on: https://go-review.googlesource.com/134635


Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Carlos Eduardo Seo <cseo@linux.vnet.ibm.com>
Reviewed-by: Michael Munday <mike.munday@ibm.com>