crypto/sha512: reduce add usage on PPC64

crypto/sha512: reduce add usage on PPC64

Similar to sha256, minimize add usage by preloading
constants. This results in a small performance uplift.

Likewise, cleanup some unused macros and registers to
make room for constants.

On ppc64le/power9:

Hash8Bytes/New     22.7MB/s ± 0%  24.1MB/s ± 0%  +6.49%
Hash8Bytes/Sum384  23.4MB/s ± 0%  24.9MB/s ± 0%  +6.32%
Hash8Bytes/Sum512  23.5MB/s ± 0%  24.9MB/s ± 0%  +6.18%
Hash1K/New          422MB/s ± 0%   455MB/s ± 0%  +7.92%
Hash1K/Sum384       424MB/s ± 0%   457MB/s ± 0%  +7.78%
Hash1K/Sum512       424MB/s ± 0%   457MB/s ± 0%  +7.77%
Hash8K/New          488MB/s ± 0%   528MB/s ± 0%  +8.18%
Hash8K/Sum384       481MB/s ± 0%   528MB/s ± 0%  +9.76%
Hash8K/Sum512       488MB/s ± 0%   515MB/s ± 0%  +5.60%

Change-Id: Ic604b482e3f6a9680b89c71399f85442f06fef3f
Reviewed-on: https://go-review.googlesource.com/c/go/+/460459
Reviewed-by: Archana Ravindar <aravind5@in.ibm.com>
Run-TryBot: Carlos Eduardo Seo <carlos.seo@linaro.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Carlos Eduardo Seo <carlos.seo@linaro.org>
Reviewed-by: Lynn Boger <laboger@linux.vnet.ibm.com>
Reviewed-by: Bryan Mills <bcmills@google.com>