fix etcd lease

[helinwang]

I made a comment in WuYi's PR that this is not necessary, so WuYi
removed it. Turns out it's necessary after confirming with coreOS
developer.

[typhoonzero]

LGTM. I'll try to figure out why we need this today.

[typhoonzero]

The steps are:

ch := make(chan *LeaseKeepAliveResponse, leaseResponseChSize)

and then start a go routine to loop and send lease grant request every TTL:

	l.firstKeepAliveOnce.Do(func() {
		go l.recvKeepAliveLoop()
		go l.deadlineLoop()
	})

and then write the response to the channel

// send update to all channels
	nextKeepAlive := time.Now().Add((time.Duration(karesp.TTL) * time.Second) / 3.0)
	ka.deadline = time.Now().Add(time.Duration(karesp.TTL) * time.Second)
	for _, ch := range ka.chs {
		select {
		case ch <- karesp:
			ka.nextKeepAlive = nextKeepAlive
		default:
		}

You can see the channel is writed using select so it's non-blocking write. So TTL will still refresh as you wanted, but if you don't consume the messages in the channel, etcd client will cosume more and more memory as it runs.

[helinwang]

@typhoonzero The channel buffer will be capped at leaseResponseChSize, so it looks fine to not eat all the messages. I would rather not do it, if not necessary, since it adds more code (complexity).
I have created a issue hoping to get an authoritative answer: etcd-io/etcd#8168

[helinwang]

@typhoonzero I dig a littler more into the code as well. Seems if we don't eat the message, TTL will be send to etcd every 500ms instead of TTL time: etcd-io/etcd#7446 It's an area that I think the etcd document could improve.