Rework outbound buffer management

Outbound element buffers now come from the sing allocator sized to the
actual packet instead of the bounded MaxMessageSize pool, element and
container pools become plain sync.Pools, and the bounded message buffer
pool serves only the receive path. Packets injected via
InputPacket/InputPackets are dropped before they are copied once a peer
has 2048 packets queued: injection runs on the caller's read loop, which
must never block on pool exhaustion, and the queues are bounded in
containers, so a flood was buffered instead of dropped.

device/peer.go

@@ -25,6 +25,8 @@ type Peer struct {
 	rxBytes           atomic.Uint64  // bytes received from peer
 	lastHandshakeNano atomic.Int64   // nano seconds since epoch
 
+	queuedOutboundPackets atomic.Int32 // packets in staged+outbound queues, for input backpressure
+
 	endpoint struct {
 		sync.Mutex
 		val            conn.Endpoint
@@ -193,6 +195,7 @@ func (peer *Peer) Start() {
 	// reset routine state
 	peer.stopping.Wait()
 	peer.stopping.Add(2)
+	peer.queuedOutboundPackets.Store(0)
 
 	peer.handshake.mutex.Lock()
 	peer.handshake.lastSentHandshake = time.Now().Add(-(RekeyTimeout + time.Second))