bounds check elimation in main loop of encryption

[xtaci]

Before

$ go build -gcflags="-d=ssa/check_bce/debug=1"  sm4.go
# command-line-arguments
./sm4.go:171:8: Found IsInBounds
./sm4.go:171:23: Found IsInBounds
./sm4.go:171:52: Found IsInBounds
./sm4.go:172:17: Found IsInBounds
./sm4.go:172:47: Found IsInBounds
./sm4.go:178:10: Found IsInBounds
./sm4.go:178:23: Found IsInBounds
./sm4.go:179:12: Found IsInBounds
./sm4.go:180:12: Found IsInBounds
./sm4.go:181:12: Found IsInBounds
./sm4.go:191:9: Found IsInBounds
./sm4.go:191:16: Found IsInBounds
./sm4.go:191:23: Found IsInBounds
./sm4.go:191:36: Found IsInBounds
./sm4.go:193:36: Found IsInBounds
./sm4.go:195:36: Found IsInBounds
./sm4.go:197:36: Found IsInBounds
./sm4.go:202:9: Found IsInBounds
./sm4.go:202:16: Found IsInBounds
./sm4.go:202:23: Found IsInBounds
./sm4.go:202:36: Found IsInBounds
./sm4.go:204:36: Found IsInBounds
./sm4.go:206:36: Found IsInBounds
./sm4.go:208:36: Found IsInBounds
./sm4.go:212:28: Found IsInBounds
./sm4.go:212:34: Found IsInBounds
./sm4.go:226:14: Found IsInBounds

After

$ go build -gcflags="-d=ssa/check_bce/debug=1"  sm4.go
# command-line-arguments
./sm4.go:171:8: Found IsInBounds
./sm4.go:171:23: Found IsInBounds
./sm4.go:171:52: Found IsInBounds
./sm4.go:172:17: Found IsInBounds
./sm4.go:172:47: Found IsInBounds
./sm4.go:178:10: Found IsInBounds
./sm4.go:178:23: Found IsInBounds
./sm4.go:179:12: Found IsInBounds
./sm4.go:180:12: Found IsInBounds
./sm4.go:181:12: Found IsInBounds
./sm4.go:192:7: Found IsInBounds
./sm4.go:195:16: Found IsSliceInBounds
./sm4.go:207:16: Found IsSliceInBounds
./sm4.go:232:14: Found IsInBounds

benchmark via kcp-go shows that we get 6~10% performance gain in this improvement
Before:

$ go test -cpuprofile=cpu.out -run=^$ -bench SM
beginning tests, encryption:salsa20, fec:10/3
goos: darwin
goarch: amd64
pkg: github.com/xtaci/kcp-go
BenchmarkSM4-4   	   50000	     33991 ns/op	  88.26 MB/s	       0 B/op	       0 allocs/op
PASS
ok  	github.com/xtaci/kcp-go	2.287s

After

$ go test -cpuprofile=cpu.out -run=^$ -bench SM
beginning tests, encryption:salsa20, fec:10/3
goos: darwin
goarch: amd64
pkg: github.com/xtaci/kcp-go
BenchmarkSM4-4   	   50000	     31625 ns/op	  94.86 MB/s	       0 B/op	       0 allocs/op
PASS
ok  	github.com/xtaci/kcp-go	2.222s