Explicitly respond with an error when cannot find an agent

client-go (Go Kubernetes client) sends a timeout value in the URL query parameter (e.g. timeout=32s). kas can parse that and close to the timeout (e.g. after 95% of time), right before the request expires, respond with an error, saying that there was a timeout finding a suitable connected agent (or any other conditions that we can detect, not sure what else).

More ideas to improve error message:

• Show the number of connected agent instances (i.e. how many Pods with agent id = X are there).
• If the above is non zero, show the number of Pods that don't have a gRPC method we are looking for.