Major optimization to the safe version of xxhash on Go 1.7 (tip)

Based on https://docs.google.com/document/d/1vdAEAjYdzjnPA9WDOQ1e4e05cYVMpqSxJYZT33Cqw2g/edit

tested with https://docs.google.com/document/d/1vdAEAjYdzjnPA9WDOQ1e4e05cYVMpqSxJYZT33Cqw2g/edit#

	name                    old time/op    new time/op    delta
	XXChecksum32-8          1.35µs ±12%    0.67µs ± 3%  -50.49%          (p=0.008 n=5+5)
	XXChecksumString32-8    1.78µs ± 2%    1.21µs ± 1%  -32.06%          (p=0.008 n=5+5)
	XXChecksum64-8          1.14µs ± 5%    0.35µs ± 2%  -69.70%          (p=0.008 n=5+5)
	XXChecksumString64-8    1.68µs ± 1%    0.86µs ± 1%  -48.74%          (p=0.008 n=5+5)

native/xxhash_safe.go

@@ -14,12 +14,14 @@ func newbyteReader(in []byte) byteReader {
 }
 
 func (br byteReader) Uint32(i int) uint32 {
-	return uint32(br[i]) | uint32(br[i+1])<<8 | uint32(br[i+2])<<16 | uint32(br[i+3])<<24
+	br = br[i : i+4]
+	return uint32(br[0]) | uint32(br[1])<<8 | uint32(br[2])<<16 | uint32(br[3])<<24
 }
 
 func (br byteReader) Uint64(i int) uint64 {
-	return uint64(br[i]) | uint64(br[i+1])<<8 | uint64(br[i+2])<<16 | uint64(br[i+3])<<24 |
-		uint64(br[i+4])<<32 | uint64(br[i+5])<<40 | uint64(br[i+6])<<48 | uint64(br[i+7])<<56
+	br = br[i : i+8]
+	return uint64(br[0]) | uint64(br[1])<<8 | uint64(br[2])<<16 | uint64(br[3])<<24 |
+		uint64(br[4])<<32 | uint64(br[5])<<40 | uint64(br[6])<<48 | uint64(br[7])<<56
 }
 
 func (br byteReader) Byte(i int) byte {