Title: Support for resource readiness feedback in XDS.

Description:
We are experiencing temporary 503 responses from Envoy when adding a new route. We managed to link those to a scenario where the new route also introduces a new cluster. When the new cluster is added it enters the warming state. As we send the routes immediately after sending clusters, there is a short period of time when requests are being routed to the cluster that is still in the warming state. As a consequence, 503 responses are returned.

We understand we don't respect following requirement from [1]: "management plane must ensure that clusters referenced by a route are in place, before pushing the updates for a route.", but it's not clear to us what is the correct protocol for determining if "clusters referenced by a route are in place".

In our experiments it looks like XDS ACKs to cluster update messages were sent from Envoy before cluster warming period is done, making them unsuitable for the purpose of detecting cluster readiness. Can we get a confirmation/refutation of this behavior?

If this is indeed the case, we would like to propose an extension of the XDS protocol to support a readiness feedback for the resources that require warming up (e.g. CDS and LDS).

[1] https://github.com/envoyproxy/data-plane-api/blob/master/XDS_PROTOCOL.md#eventual-consistency-considerations