Add support for CPU-specific optimisations

[pjbgf]

The cgo version of sha1cd currently servers two purposes:

• Provides an easy way to assure 100% fidelity with the upstream implementation.
• Provides best performance for larger payloads.

Once SIMD is implemented, use cases such as go-git can solely rely on the Pure Go implementation and no longer have a requirement to build using CGO_ENABLED=1.

[pjbgf]

The performance improvements introduced in the generic implemementation (#34) should make an impact across all platforms. In arm64 this is the delta:

goos: linux
goarch: arm64
pkg: github.com/pjbgf/sha1cd/test
                               │       before       │                    after                    │
                               │       sec/op       │       sec/op         vs base                │
Hash8Bytes/sha1cd-4                   23.050µ ±  3%         3.526µ ±   5%  -84.70% (p=0.000 n=10)
Hash320Bytes/sha1cd-4                 111.24µ ±  3%         19.60µ ±   4%  -82.38% (p=0.000 n=10)
Hash1K/sha1cd-4                       293.45µ ±  2%         53.37µ ±   4%  -81.81% (p=0.000 n=10)
Hash8K/sha1cd-4                       2191.2µ ±  3%         402.3µ ±   5%  -81.64% (p=0.000 n=10)
HashWithCollision/sha1cd-4                                  61.41µ ±   4%


                               │     before      │             after            │
                               │       B/s       │      B/s        vs base      │
Hash8Bytes/sha1cd-4                336.9Ki ±  1%   2216.8Ki ±  4%  +557.97% (p=0.000 n=10)
Hash320Bytes/sha1cd-4              2.742Mi ±  3%   15.569Mi ±  3%  +467.83% (p=0.000 n=10)
Hash1K/sha1cd-4                    3.328Mi ±  2%   18.296Mi ±  4%  +449.71% (p=0.000 n=10)
Hash8K/sha1-4                      27.26Mi ±  2%    27.17Mi ±  1%         ~ (p=0.956 n=10)
HashWithCollision/sha1cd-4                          9.937Mi ±  4%

Making the generic version good enough to be the default, as it beats the cgo implementation for some payload sizes:

BenchmarkHash8Bytes/sha1-4     	                  470190	      2550 ns/op	   3.14 MB/s	       0 B/op	       0 allocs/op
BenchmarkHash8Bytes/sha1cd-4      	          345133	      3504 ns/op	   2.28 MB/s	       0 B/op	       0 allocs/op
BenchmarkHash8Bytes/sha1cd_cgo-4         	   84704	     14858 ns/op	   0.54 MB/s	    2688 B/op	       1 allocs/op
BenchmarkHash320Bytes/sha1-4             	   78384	     13745 ns/op	  23.28 MB/s	       0 B/op	       0 allocs/op
BenchmarkHash320Bytes/sha1cd-4           	   63459	     19532 ns/op	  16.38 MB/s	       0 B/op	       0 allocs/op
BenchmarkHash320Bytes/sha1cd_cgo-4       	   51583	     25543 ns/op	  12.53 MB/s	    2688 B/op	       1 allocs/op
BenchmarkHash1K/sha1-4                   	   31378	     38622 ns/op	  26.51 MB/s	       0 B/op	       0 allocs/op
BenchmarkHash1K/sha1cd-4                 	   22035	     53162 ns/op	  19.26 MB/s	       0 B/op	       0 allocs/op
BenchmarkHash1K/sha1cd_cgo-4             	   29412	     39308 ns/op	  26.05 MB/s	    2688 B/op	       1 allocs/op
BenchmarkHash8K/sha1-4                   	    4258	    293192 ns/op	  27.94 MB/s	       0 B/op	       0 allocs/op
BenchmarkHash8K/sha1cd-4                 	    3087	    391253 ns/op	  20.94 MB/s	       0 B/op	       0 allocs/op
BenchmarkHash8K/sha1cd_cgo-4             	    5084	    211308 ns/op	  38.77 MB/s	    2688 B/op	       1 allocs/op
BenchmarkHashWithCollision/sha1cd-4      	   18715	     62284 ns/op	  10.28 MB/s	       0 B/op	       0 allocs/op
BenchmarkHashWithCollision/sha1cd_cgo-4  	   28069	     44771 ns/op	  14.29 MB/s	    2688 B/op	       1 allocs/op