Not able to use non-UTF-8 strings as keys

[limbulvetr]

Hello there. I've been integrating my C++ project into golang. I'm unable to generate decoder and unmarshaler because the project uses one-byte keys in msgpack to shorten the payload.

type ValueSegment struct {
	Offset int    `msg:"\x0b"`
	...
}

According to spec, strings such as "\x0b" are allowed as keys.

tools/imports fails to format the generated code because it contains illegal characters which should have been escaped. The fix is trivial.

--- a/src/github.com/tinylib/msgp/gen/decode.go
+++ b/src/github.com/tinylib/msgp/gen/decode.go
@@ -102,7 +102,7 @@ func (d *decodeGen) structAsMap(s *Struct) {
 	d.p.print("\nswitch msgp.UnsafeString(field) {")
 	for i := range s.Fields {
 		d.ctx.PushString(s.Fields[i].FieldName)
-		d.p.printf("\ncase \"%s\":", s.Fields[i].FieldTag)
+		d.p.printf("\ncase %q:", s.Fields[i].FieldTag)
 		next(d, s.Fields[i].FieldElem)
 		d.ctx.Pop()
 		if !d.p.ok() {

--- a/src/github.com/tinylib/msgp/gen/unmarshal.go
+++ b/src/github.com/tinylib/msgp/gen/unmarshal.go
@@ -105,7 +105,7 @@ func (u *unmarshalGen) mapstruct(s *Struct) {
 		if !u.p.ok() {
 			return
 		}
-		u.p.printf("\ncase \"%s\":", s.Fields[i].FieldTag)
+		u.p.printf("\ncase %q:", s.Fields[i].FieldTag)
 		u.ctx.PushString(s.Fields[i].FieldName)
 		next(u, s.Fields[i].FieldElem)
 		u.ctx.Pop()

[philhofer]

The fix looks fine to me; feel like sending a pull request? Thanks.