- improve xor.go performance by re-organizing code layout.
- dramatically reduce zero-ing operations in FEC.
$ go version
go version go1.7.3 darwin/amd64
xtaci/smux@ee8b5b5
xtaci/kcp-go@7112c1c
4ccc922
MD5 (kcptun-linux-amd64-20161102.tar.gz) = 701fcff24a5485384bf38d196b2217b6
MD5 (kcptun-linux-386-20161102.tar.gz) = 56becb5632919082bf8b40b3358dc58c
MD5 (kcptun-darwin-amd64-20161102.tar.gz) = a8d7d7f5e6daa8ef9d4fcc38705e4535
MD5 (kcptun-darwin-386-20161102.tar.gz) = 8aa3f1a6c04f83db8372b0de9272d80f
MD5 (kcptun-windows-amd64-20161102.tar.gz) = b2b3280451c311973b65d037cee7a426
MD5 (kcptun-windows-386-20161102.tar.gz) = dabc8e8cc275c574963131a4143745c4
MD5 (kcptun-freebsd-amd64-20161102.tar.gz) = e23a9d4eea3da8721c22ab37adc50193
MD5 (kcptun-freebsd-386-20161102.tar.gz) = da8633e73e6ec390669c8e5b2d8d6020
MD5 (kcptun-linux-arm-20161102.tar.gz) = 94adbfd1bb2d7dafa1dd4ff8ea51c949