WorkerPool labels take long time to propagate to Worker resources

When the `metadata.labels` field of WorkerPool changes, it takes a long time (5 minutes) for the change to propagate to the associated `Worker` resources.

The simplest way to reproduce using the counter demo is just by running `kubectl label workerpool counter -n ate-demo-counter repro=test` and then polling the worker until the label come up by doing `kubectl ate get worker -o yaml`.

I believe this is because the reconciler is watching pod changes, not workerpool changes, and the pod informer has a resync period of 5 minutes.

@sfunkenhauser