Perf: Configurable Broker implementation

[nachocano]

Problem
As of now, the Broker ingress and filter pods are k8s Deployments. When stress testing them, they keep on restarting due to liveness checks failures. Upon high traffic, we need to be able to scale these out.
We should be able to configure the underlying resources that a broker will use. @grantr suggested an annotation that enables us to either create Deployments or Knative Services. I agree with that approach.
This will have implications in metrics, and probably tracing as well.

Persona:
Which persona is this feature for?
Operator, Developer

Exit Criteria
Being able to configure whether to use Deployments or Knative Services as the underlying resources for Brokers.

Time Estimate (optional):
How many developer-days do you think this may take to resolve?
5 days

[n3wscott]

/close
We have moved on to optional installable broker implementations. No config needed, install the Knative Serving based broker, if it existed.

[knative-prow-robot]

@n3wscott: Closing this issue.

In response to this:

/close
We have moved on to optional installable broker implementations. No config needed, install the Knative Serving based broker, if it existed.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.