Benchmark findcipher and minor optimization

[fortuna]

Example run:

$ go test -cpuprofile cpu.prof -memprofile mem.prof -bench . -benchmem -run=^$ github.com/Jigsaw-Code/outline-ss-server/shadowsocks
goos: linux
goarch: amd64
pkg: github.com/Jigsaw-Code/outline-ss-server/shadowsocks
BenchmarkTCPFindCipher-12    	    1000	   2397037 ns/op	 2026516 B/op	    3013 allocs/op
BenchmarkUDPUnpack-12        	    2000	    892407 ns/op	  125029 B/op	    1701 allocs/op

I'm getting ~2ms to go over 100 ciphers. It makes 3013 allocations for a total of 2MB.

With 1 cipher:

BenchmarkTCPFindCipher-12    	   20000	     63841 ns/op	   20451 B/op	      33 allocs/op
BenchmarkUDPUnpack-12        	  100000	     12803 ns/op	    1280 B/op	      18 allocs/op

To inspect the details of the profile:

go tool pprof mem.prof

(then type web)

[trevj]

Cool!