fix: Adjust notifyCloseConnection to be a buffered channel to avoid blocking during exception handling

[nengwu765]

When closing the connection, an EOF error in the read method may trigger the shutdown operation. If notifyCloseConnection is unbuffered at this time, it will block graceful shutdown.

[nengwu765]

func (c *ConnectionWrapper) handleConnectionClose() {
	for {
		c.logger.Debug("handleConnectionClose is waiting for c.connected", nil)
		<-c.connected
		c.logger.Debug("handleConnectionClose is for connection or Pub/Sub close", nil)

		notifyCloseConnection := c.amqpConnection.NotifyClose(make(chan *amqp.Error))

		select {
		case <-c.closing:
			c.logger.Debug("Stopping handleConnectionClose", nil)
			c.connected = make(chan struct{})
			return
		case err := <-notifyCloseConnection:
			c.connected = make(chan struct{})
			c.logger.Error("Received close notification from AMQP, reconnecting", err, nil)
			c.reconnect()
		}
	}
}

when close happening, case <-c.closing done. notifyCloseConnection become a none consume channel。other gorutine send err to this channel will be blocked.