daemon/peer: try reconnecting based on bitcoind poll time, not 15 seconds.

We seem to be getting spurious reconnect failures on Travis, this
should fix it (15 seconds may be too long at worst case).

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

daemon/peer.c

@@ -4453,7 +4453,8 @@ static void reconnect_failed(struct io_conn *conn, struct peer *peer)
 	}
 
 	log_debug(peer->log, "Setting timer to re-connect");
-	new_reltimer(&peer->dstate->timers, peer, time_from_sec(15), try_reconnect, peer);
+	new_reltimer(&peer->dstate->timers, peer, peer->dstate->config.poll_time,
+		     try_reconnect, peer);
 }
 
 static struct io_plan *init_conn(struct io_conn *conn, struct peer *peer)