speed up marshaling of transport parameters (#3531)

The speedup comes from multiple sources:
1. We now preallocate a byte slice, instead of appending multiple times.
2. Marshaling into a byte slice is faster than using a bytes.Buffer.
3. quicvarint.Write allocates, while quicvarint.Append doesn't.

fuzzing/transportparameters/cmd/corpus.go

@@ -1,7 +1,6 @@
 package main
 
 import (
-	"bytes"
 	"log"
 	"math"
 	"math/rand"
@@ -78,9 +77,7 @@ func main() {
 			}
 			data = tp.Marshal(pers)
 		} else {
-			b := &bytes.Buffer{}
-			tp.MarshalForSessionTicket(b)
-			data = b.Bytes()
+			data = tp.MarshalForSessionTicket(nil)
 		}
 		if err := helper.WriteCorpusFileWithPrefix("corpus", data, transportparameters.PrefixLen); err != nil {
 			log.Fatal(err)

fuzzing/transportparameters/fuzz.go

@@ -51,10 +51,9 @@ func fuzzTransportParametersForSessionTicket(data []byte) int {
 	if err := tp.UnmarshalFromSessionTicket(bytes.NewReader(data)); err != nil {
 		return 0
 	}
-	buf := &bytes.Buffer{}
-	tp.MarshalForSessionTicket(buf)
+	b := tp.MarshalForSessionTicket(nil)
 	tp2 := &wire.TransportParameters{}
-	if err := tp2.UnmarshalFromSessionTicket(bytes.NewReader(buf.Bytes())); err != nil {
+	if err := tp2.UnmarshalFromSessionTicket(bytes.NewReader(b)); err != nil {
 		panic(err)
 	}
 	return 1