Manually unlock the mutex in ClientMap.SendQueue.

Rather than use defer. It is only a tiny amount faster, but this
function is frequently called.

Before:
	$ go test -bench=BenchmarkSendQueue -benchtime=2s
	BenchmarkSendQueue-4    15901834               151 ns/op
After:
	$ go test -bench=BenchmarkSendQueue -benchtime=2s
	BenchmarkSendQueue-4    15859948               147 ns/op

https://gitlab.torproject.org/tpo/anti-censorship/pluggable-transports/snowflake/-/issues/40177

common/turbotunnel/clientmap.go

@@ -57,8 +57,9 @@ func NewClientMap(timeout time.Duration) *ClientMap {
 // necessary.
 func (m *ClientMap) SendQueue(addr net.Addr) chan []byte {
 	m.lock.Lock()
-	defer m.lock.Unlock()
-	return m.inner.SendQueue(addr, time.Now())
+	queue := m.inner.SendQueue(addr, time.Now())
+	m.lock.Unlock()
+	return queue
 }
 
 // clientMapInner is the inner type of ClientMap, implementing heap.Interface.

common/turbotunnel/clientmap_test.go

@@ -0,0 +1,18 @@
+package turbotunnel
+
+import (
+	"testing"
+	"time"
+)
+
+// Benchmark the ClientMap.SendQueue function. This is mainly measuring the cost
+// of the mutex operations around the call to clientMapInner.SendQueue.
+func BenchmarkSendQueue(b *testing.B) {
+	m := NewClientMap(1 * time.Hour)
+	id := NewClientID()
+	m.SendQueue(id) // populate the entry for id
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		m.SendQueue(id)
+	}
+}