Update replicator to use a timeout and to automatically stop watching when a final phase is seen.

[bunchesofdonald]

Digging into issue #55 I found that the thread hanging was caused by kubernetes.watch.Watch not obeying the stop message. This happens because stop is only evaluated after watch.stream yields and it will stop yielding when the job we're watching is complete. This PR makes two adjustments to the way we're handling the watcher:

1. It passes in a timeout of pod_watch_timeout_seconds to watch.stream which means that if any given request that watch makes times out then our thread will exit.
2. In the event processing loop watch.stop is explicitly called whenever a 'final' phase is seen.

These two changes should ensure that the thread exits at most pod_watch_timeout_seconds seconds after the last event is seen, and in most cases the thread should exit immediately after a flow-run is complete.

Closes #55

Checklist

• References any related issue by including "Closes #" or "Closes ".
  • If no issue exists and your change is not a small fix, please create an issue first.
• Includes tests or only affects documentation.
• Passes pre-commit checks.
  • Run pre-commit install && pre-commit run --all locally for formatting and linting.
• Summarizes PR's changes in CHANGELOG.md

[desertaxle]

LGTM!