Benchmark FastXOR on non amd64 platforms

[nhooyr]

golang/go#31586 (comment)

It might be slower.

[nhooyr]

It's unlikely to be any slower and ARM also supports unaligned loads/stores and is little endian so I don't expect there to be any issues here given ARM and AMD64 are the major architectures in use.

If anyone finds it too slow, they can open a new issue.

[nhooyr]

I will test to confirm on an AWS arm instance.

[nhooyr]

Tested on an AWS a1.xlarge aarch64 ARM instance.

ubuntu@ip-172-31-31-187:~/websocket$ go test -run=^\$ -bench=BenchmarkXOR
goos: linux
goarch: arm64
pkg: nhooyr.io/websocket
BenchmarkXOR/2/basic-4         	100000000	        10.0 ns/op	 199.39 MB/s	       0 B/op	       0 allocs/op
BenchmarkXOR/2/fast-4          	100000000	        12.7 ns/op	 158.10 MB/s	       0 B/op	       0 allocs/op
BenchmarkXOR/16/basic-4        	50000000	        34.5 ns/op	 464.38 MB/s	       0 B/op	       0 allocs/op
BenchmarkXOR/16/fast-4         	50000000	        26.6 ns/op	 601.40 MB/s	       0 B/op	       0 allocs/op
BenchmarkXOR/32/basic-4        	20000000	        69.6 ns/op	 459.61 MB/s	       0 B/op	       0 allocs/op
BenchmarkXOR/32/fast-4         	50000000	        30.5 ns/op	1048.10 MB/s	       0 B/op	       0 allocs/op
BenchmarkXOR/512/basic-4       	 2000000	       906 ns/op	 565.00 MB/s	       0 B/op	       0 allocs/op
BenchmarkXOR/512/fast-4        	20000000	       102 ns/op	4973.82 MB/s	       0 B/op	       0 allocs/op
BenchmarkXOR/4096/basic-4      	  200000	      7157 ns/op	 572.23 MB/s	       0 B/op	       0 allocs/op
BenchmarkXOR/4096/fast-4       	 2000000	       660 ns/op	6197.44 MB/s	       0 B/op	       0 allocs/op
BenchmarkXOR/16384/basic-4     	   50000	     28607 ns/op	 572.71 MB/s	       0 B/op	       0 allocs/op
BenchmarkXOR/16384/fast-4      	  500000	      2631 ns/op	6225.87 MB/s	       0 B/op	       0 allocs/op
PASS
ok  	nhooyr.io/websocket	19.901s

Still faster than basic. Not as much as amd64 but good enough to keep as is.