Initial queue-proxy websocket errors are confusing.

[markusthoemmes]

Expected Behavior

The queue-proxy does not report any errors after startup if it can successfully connect to the autoscaler via a websocket connection.

Actual Behavior

The queue-proxy almost always reports one or two failures because an event was tried to be sent before an initial connection was created.

Steps to Reproduce the Problem

1. Deploy any application.
2. Check the queue-proxy logs for "connection has not yet been established"

Additional Info

To mitigate this, we could only start reporting these kinds of errors a certain time after startup. 10 seconds should suffice in this case I think.

[markusthoemmes]

/assign

[owenthereal]

I updated on #2892 (comment) that the readiness probe for queue-proxy failed and caused the pod to be terminated. Not sure if it's related.