fix: shutdown the tcpstream when we initiate to close a peer connection

[garyyu]

Based on #2287 's a test log, perhaps I find one leaking point for peer connections:

20190104 05:15:29.510 DEBUG grin_p2p::peer - Send tx kernel hash 133cc0ab to 47.99.92.164:13414
20190104 05:15:33.058 DEBUG grin_p2p::peer - Client 47.99.92.164:13414 connection lost: Connection(Custom { kind: ConnectionAborted, error: StringError("read_exact") })
20190104 05:15:50.546 DEBUG grin_p2p::peers - clean_peers V4(47.99.92.164:13414), not connected
20190104 05:22:28.466 DEBUG grin_p2p::serv - clean_lost_sockets - close a lost open socket: 47.99.92.164:58462, not in peers list

that means when we quit the peer polling thread, we forgot to close the socket:

	// check the close channel
	if let Ok(_) = close_rx.try_recv() {
		debug!(
			"Connection close with {} initiated by us",
			...
		);
		break;    <<<< problem here!  we just quit without a real close!
	}