Fix autoscaling bug with number of replicas

[edavidj]

What this PR does / why we need it?:

• Adds a guard to the deployment/statefulset customization functions that checks for the existence of autoscaler in the spec before updating the # of replicas in the resource. This prevents the loop of operator contradicting the HPA.
• My changes assume that we do not want the operator to touch the # of replicas if there's an autoscaler enabled.
• Updated the autoscaler test to watch that the deployment actually gets to the minimum number of replicas before passing the test. Rather than just watching the HPA resource definition.
  • Note that I actually don't think this e2e CAN catch this specific failure, as it still technically reaches 2 replicas momentarily and then starts going up and down over and over. This would be better mandated by unit tests.

Which issue(s) this PR fixes:

Fixes #68

[leochr]

Looks good. Thanks @edavidj!