Issue #278: optimize normalizeHeaderKey a bit

Performance results on amd64:

name                            old time/op  new time/op  delta
NormalizeHeaderKeyCommonCase-4  43.3ns ± 0%  32.6ns ± 1%  -24.80%  (p=0.000 n=9+10)
NormalizeHeaderKeyLowercase-4   42.6ns ± 3%  32.6ns ± 1%  -23.41%  (p=0.000 n=10+10)
NormalizeHeaderKeyUppercase-4   43.5ns ± 1%  32.6ns ± 2%  -25.03%  (p=0.000 n=9+8)

Based on top of https://github.com/valyala/fasthttp/pull/279 .

bytesconv.go

@@ -343,18 +343,34 @@ func hexbyte2int(c byte) int {
 
 const toLower = 'a' - 'A'
 
-func uppercaseByte(p *byte) {
-	c := *p
-	if c >= 'a' && c <= 'z' {
-		*p = c - toLower
+var toLowerTable = func() [256]byte {
+	var a [256]byte
+	for i := 0; i < 256; i++ {
+		c := byte(i)
+		if c >= 'A' && c <= 'Z' {
+			a[i] = toLower
+		}
 	}
+	return a
+}()
+
+var toUpperTable = func() [256]byte {
+	var a [256]byte
+	for i := 0; i < 256; i++ {
+		c := byte(i)
+		if c >= 'a' && c <= 'z' {
+			a[i] = 256 - toLower
+		}
+	}
+	return a
+}()
+
+func uppercaseByte(p *byte) {
+	*p += toUpperTable[*p]
 }
 
 func lowercaseByte(p *byte) {
-	c := *p
-	if c >= 'A' && c <= 'Z' {
-		*p = c + toLower
-	}
+	*p += toLowerTable[*p]
 }
 
 func lowercaseBytes(b []byte) {

header.go

@@ -2014,19 +2014,20 @@ func normalizeHeaderKey(b []byte, disableNormalizing bool) {
 	}
 
 	n := len(b)
-	up := true
-	for i := 0; i < n; i++ {
-		switch b[i] {
-		case '-':
-			up = true
-		default:
-			if up {
-				up = false
+	if n == 0 {
+		return
+	}
+
+	uppercaseByte(&b[0])
+	for i := 1; i < n; i++ {
+		if b[i] == '-' {
+			i++
+			if i < n {
 				uppercaseByte(&b[i])
-			} else {
-				lowercaseByte(&b[i])
 			}
+			continue
 		}
+		lowercaseByte(&b[i])
 	}
 }