Implement backoff for k8s watcher retries

[cbgbt]

Issue number:
#478

Description of changes:
The kube::runtime provides us utilities for watching for events from Kubernetes watch APIs and updating an internal cache of state, which is much more efficient than polling the k8s API for changes.

If the brupop CRD is not installed, these watchers end up in an extremely tight retry loop to begin watching.

This PR:

• Updates kube to get access to some nice default backoff functionality that was added in a recent release.
• Uses aforementioned backoff functionality
• Unrelatedly, lets the controller sleep for a longer fixed interval if there is nothing to do (which is true if the CRD is not installed)

Testing done:
Deployed the changes to a cluster without the CRD installed.

Prior to this change, the agent logs were, several times per second, spammed with messages like this:

2023-08-08T20:23:28.561457Z  WARN kube_client::client: Unsuccessful data error parse: 404 page not found

    at /src/.cargo/registry/src/index.crates.io-6f17d22bba15001f/kube-client-0.85.0/src/client/mod.rs:445

After this change, these messages are clearly backed-off exponentially, with a short initial burst followed by longer waits of 15+ seconds.

The controller also emits this log every 10 seconds:

2023-08-08T20:24:59.897841Z  INFO controller::controller: Nothing to do: The bottlerocket-update-operator is not aware of any BottlerocketShadow objects. Is the bottlerocket-shadow CRD installed? Are nodes labelled so that the agent is deployed to them? See the project's README for more information.
    at controller/src/controller.rs:285

Terms of contribution:

By submitting this pull request, I agree that this contribution is dual-licensed under the terms of both the Apache License, version 2.0, and the MIT license.

[stmcginnis]

Nice!