fix(network): deadlock in lwIP UDP handler's Close function

[jyyi1]

This PR fixes a deadlock bug when a careless PacketRequestSender tries to Close the corresponding PacketResponseReceiver multiple times, here is the buggy flow chart:

user calls PacketResponseReceiver.Close()
                  ↓
     udpConnResponseWriter.Close()  <----------------------------------------(1)
                  ↓                                                           |
     udpHandler.closeSession(conn)                                            |
                  ↓                                                           |
              h.mu.Lock()                                                     |
                  ↓                                                           |
        h.senders[laddr].Close() → PacketRequestSender.Close() → PacketResponseReceiver.Close()

Note the recursive call path (1): h.senders[laddr].Close() will block because h.mu has already be held, and it won't Unlock() until h.senders[laddr].Close() returns, this causes deadlock!

This fix simply added a closed indicator in udpConnResponseWriter to make sure only one closeSession will be called:

  user calls PacketResponseReceiver.Close()
                  ↓
     udpConnResponseWriter.Close()
                  ↓
r.closed.CompareAndSwap(false, true) == true
                  ↓
     udpHandler.closeSession(conn)                                            x
                  ↓                                                           |
              h.mu.Lock()                                r.closed.CompareAndSwap(false, true) == false
                  ↓                                                           |
        h.senders[laddr].Close() → PacketRequestSender.Close() → PacketResponseReceiver.Close()
                  ↓
             h.mu.UnLock()