v0.6.0-alpha.4: capacity controller: don't retry TargetDeploymentCountError

Whenever the capacity controller can't find the exact number of
deployments it expects, it should not keep on retrying: it's impossible
for the error to resolve if deployments aren't created or deleted to
satisfy the expectation. Since the controller also listens to
deployments being created and deleted, just waiting for those events
should be enough for a new sync of the CapacityTarget to trigger, and
the controller will check again if the expectation is satisfied.

This should reduce a lot of spurious retries when a deployment just got
created and the change hasn't propagated to the informers yet, and will
also make the resync loop much faster in case shipper has a lot of
broken releases currently active.