Detect SSE automatically and enable corresponding optimized code path.

With this change, following happens.

```
if sse {
   compressSSE()
   return
}
compressGeneric()
```

compressGeneric is used as a fallback when SSE is not detected.

blake2b.go

@@ -143,6 +143,7 @@ func (d *digest) initialize(c *Config) {
 		p[2] = 1
 		p[3] = 1
 	}
+
 	// Initialize.
 	d.size = c.Size
 	for i := 0; i < 8; i++ {
@@ -151,6 +152,7 @@ func (d *digest) initialize(c *Config) {
 	if c.Tree != nil && c.Tree.IsLastNode {
 		d.isLastNode = true
 	}
+
 	// Process key.
 	if c.Key != nil {
 		copy(d.paddedKey[:], c.Key)
@@ -231,11 +233,11 @@ func (d *digest) Write(p []byte) (nn int, err error) {
 }
 
 // Sum returns the calculated checksum.
-func (d0 *digest) Sum(in []byte) []byte {
-	// Make a copy of d0 so that caller can keep writing and summing.
-	d := *d0
-	hash := d.checkSum()
-	return append(in, hash[:d.size]...)
+func (d *digest) Sum(in []byte) []byte {
+	// Make a copy of d so that caller can keep writing and summing.
+	d0 := *d
+	hash := d0.checkSum()
+	return append(in, hash[:d0.size]...)
 }
 
 func (d *digest) checkSum() [Size]byte {