crypto/sha512: improve performance of riscv64 assembly

crypto/sha512: improve performance of riscv64 assembly

Implement optimised versions of Maj and Ch, which reduce the number of
instructions required per round. Reorder instructions for better
interleaving.

This gives around a 10% gain on a StarFive VisionFive 2:

                    │  sha512.1   │              sha512.2               │
                    │   sec/op    │   sec/op     vs base                │
Hash8Bytes/New-4      9.310µ ± 0%   8.564µ ± 0%   -8.01% (p=0.000 n=10)
Hash8Bytes/Sum384-4   8.833µ ± 0%   7.980µ ± 0%   -9.66% (p=0.000 n=10)
Hash8Bytes/Sum512-4   9.293µ ± 0%   8.162µ ± 0%  -12.17% (p=0.000 n=10)
Hash1K/New-4          49.60µ ± 0%   44.33µ ± 0%  -10.63% (p=0.000 n=10)
Hash1K/Sum384-4       48.93µ ± 0%   43.78µ ± 0%  -10.53% (p=0.000 n=10)
Hash1K/Sum512-4       49.48µ ± 0%   43.96µ ± 0%  -11.15% (p=0.000 n=10)
Hash8K/New-4          327.9µ ± 0%   292.6µ ± 0%  -10.78% (p=0.000 n=10)
Hash8K/Sum384-4       327.3µ ± 0%   292.0µ ± 0%  -10.77% (p=0.000 n=10)
Hash8K/Sum512-4       327.8µ ± 0%   292.2µ ± 0%  -10.85% (p=0.000 n=10)
geomean               52.87µ        47.31µ       -10.51%

                    │   sha512.1   │               sha512.2               │
                    │     B/s      │     B/s       vs base                │
Hash8Bytes/New-4      839.8Ki ± 0%   908.2Ki ± 0%   +8.14% (p=0.000 n=10)
Hash8Bytes/Sum384-4   888.7Ki ± 1%   976.6Ki ± 0%   +9.89% (p=0.000 n=10)
Hash8Bytes/Sum512-4   839.8Ki ± 0%   957.0Ki ± 0%  +13.95% (p=0.000 n=10)
Hash1K/New-4          19.69Mi ± 0%   22.03Mi ± 0%  +11.86% (p=0.000 n=10)
Hash1K/Sum384-4       19.96Mi ± 0%   22.31Mi ± 0%  +11.75% (p=0.000 n=10)
Hash1K/Sum512-4       19.74Mi ± 0%   22.21Mi ± 0%  +12.51% (p=0.000 n=10)
Hash8K/New-4          23.82Mi ± 0%   26.70Mi ± 0%  +12.09% (p=0.000 n=10)
Hash8K/Sum384-4       23.87Mi ± 0%   26.75Mi ± 0%  +12.07% (p=0.000 n=10)
Hash8K/Sum512-4       23.83Mi ± 0%   26.73Mi ± 0%  +12.16% (p=0.000 n=10)
geomean               7.334Mi        8.184Mi       +11.59%

Change-Id: I66e359e96b25b38efbc4d840e6b2d6a1e5d417ec
Reviewed-on: https://go-review.googlesource.com/c/go/+/605495
Reviewed-by: David Chase <drchase@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Reviewed-by: Mark Ryan <markdryan@rivosinc.com>
Reviewed-by: Meng Zhuo <mengzhuo1203@gmail.com>