crypto/md5: provide optimised assembly for riscv64

crypto/md5: provide optimised assembly for riscv64

Provide an optimised assembly implementation of MD5 for RISC-V.
There are significant performance improvements.  The assembler takes
advantage of Zbb instructions when they are available.

Results for the VisionFive 2 running Ubuntu 24.04 with
GORISCV64=rva20u64.

goos: linux
goarch: riscv64
pkg: crypto/md5
                    │ md5_go.txt  │             md5_ass.txt             │
                    │   sec/op    │   sec/op     vs base                │
Hash8Bytes            1.202µ ± 0%   1.220µ ± 0%   +1.50% (p=0.000 n=10)
Hash64                1.665µ ± 0%   1.518µ ± 0%   -8.83% (p=0.000 n=10)
Hash128               2.165µ ± 0%   1.885µ ± 0%  -12.94% (p=0.000 n=10)
Hash256               3.162µ ± 0%   2.613µ ± 0%  -17.38% (p=0.000 n=10)
Hash512               5.146µ ± 0%   4.063µ ± 0%  -21.05% (p=0.000 n=10)
Hash1K                9.115µ ± 0%   6.959µ ± 0%  -23.65% (p=0.000 n=10)
Hash8K                64.68µ ± 0%   47.52µ ± 0%  -26.54% (p=0.000 n=10)
Hash1M                8.131m ± 0%   5.936m ± 0%  -27.00% (p=0.000 n=10)
Hash8M                65.06m ± 0%   47.50m ± 0%  -26.99% (p=0.000 n=10)
Hash8BytesUnaligned   1.210µ ± 0%   1.199µ ± 0%   -0.91% (p=0.000 n=10)
Hash1KUnaligned       9.114µ ± 0%   8.266µ ± 0%   -9.30% (p=0.000 n=10)
Hash8KUnaligned       64.68µ ± 0%   57.97µ ± 0%  -10.38% (p=0.000 n=10)
geomean               22.37µ        18.83µ       -15.82%

Results for the VisionFive 2 running Ubuntu 24.04 with
GORISCV64=rva22u64.

goos: linux
goarch: riscv64
pkg: crypto/md5
                    │ md5_g22.txt │             md5_a22.txt             │
                    │   sec/op    │   sec/op     vs base                │
Hash8Bytes            1.175µ ± 0%   1.002µ ± 0%  -14.72% (p=0.000 n=10)
Hash64                1.575µ ± 0%   1.274µ ± 0%  -19.11% (p=0.000 n=10)
Hash128               2.033µ ± 0%   1.587µ ± 0%  -21.92% (p=0.000 n=10)
Hash256               2.943µ ± 0%   2.209µ ± 0%  -24.93% (p=0.000 n=10)
Hash512               4.755µ ± 0%   3.443µ ± 0%  -27.58% (p=0.000 n=10)
Hash1K                8.378µ ± 0%   5.910µ ± 0%  -29.46% (p=0.000 n=10)
Hash8K                59.12µ ± 0%   40.45µ ± 0%  -31.58% (p=0.000 n=10)
Hash1M                7.426m ± 0%   5.056m ± 0%  -31.92% (p=0.000 n=10)
Hash8M                59.41m ± 0%   40.45m ± 0%  -31.91% (p=0.000 n=10)
Hash8BytesUnaligned   1.169µ ± 0%   1.012µ ± 0%  -13.43% (p=0.000 n=10)
Hash1KUnaligned       8.379µ ± 0%   7.213µ ± 0%  -13.91% (p=0.000 n=10)
Hash8KUnaligned       59.12µ ± 0%   50.90µ ± 0%  -13.91% (p=0.000 n=10)
geomean               20.83µ        15.99µ       -23.21%

Change-Id: I61e3fa802c2cc50e0b5f71f151b4741691ccb481
Reviewed-on: https://go-review.googlesource.com/c/go/+/527936
Reviewed-by: Joel Sing <joel@sing.id.au>
Auto-Submit: Tim King <taking@google.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Reviewed-by: Meng Zhuo <mengzhuo1203@gmail.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Tim King <taking@google.com>