device: get rid of nonce routine

This moves to a simple queue with no routine processing it, to reduce
scheduler pressure.

This splits latency in half!

benchmark                  old ns/op     new ns/op     delta
BenchmarkThroughput-16     2394          2364          -1.25%
BenchmarkLatency-16        259652        120810        -53.47%

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

device/uapi.go

@@ -156,6 +156,7 @@ func (device *Device) IpcSetOperation(r io.Reader) (err error) {
 			if deviceConfig {
 				deviceConfig = false
 			}
+			peer.handlePostConfig()
 			// Load/create the peer we are now configuring.
 			err := device.handlePublicKeyLine(peer, value)
 			if err != nil {
@@ -174,6 +175,7 @@ func (device *Device) IpcSetOperation(r io.Reader) (err error) {
 			return err
 		}
 	}
+	peer.handlePostConfig()
 
 	if err := scanner.Err(); err != nil {
 		return ipcErrorf(ipc.IpcErrorIO, "failed to read input: %w", err)
@@ -241,6 +243,12 @@ type ipcSetPeer struct {
 	created bool // new reports whether this is a newly created peer
 }
 
+func (peer *ipcSetPeer) handlePostConfig() {
+	if peer.Peer != nil && !peer.dummy && peer.Peer.device.isUp.Get() {
+		peer.SendStagedPackets()
+	}
+}
+
 func (device *Device) handlePublicKeyLine(peer *ipcSetPeer, value string) error {
 	// Load/create the peer we are configuring.
 	var publicKey NoisePublicKey