Workaround for HPA dry-run metrics duplication

[stefanprodan]

This PR implements a workaround for a Kubernetes API server-side apply dry-run bug where it duplicates the first item in the metrics array and inserts an empty metric as the last item in the array.

Given:

apiVersion: autoscaling/v2beta2
kind: HorizontalPodAutoscaler
metadata:
  name: frontend
  namespace: default
spec:
  behavior:
    scaleDown:
      policies:
        - periodSeconds: 15
          type: Percent
          value: 100
      selectPolicy: Max
      stabilizationWindowSeconds: 60
    scaleUp:
      policies:
        - periodSeconds: 15
          type: Pods
          value: 4
        - periodSeconds: 15
          type: Percent
          value: 100
      selectPolicy: Max
      stabilizationWindowSeconds: 0
  maxReplicas: 30
  metrics:
    - type: Pods
      pods:
        metric:
          name: nginx_http_requests_total
        target:
          averageValue: "4"
          type: AverageValue
  minReplicas: 2
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: frontend

The dry-run diff returns bogus data:

$ kubectl diff --server-side -f hpa.yaml
@@ -74,6 +97,13 @@
         averageValue: "4"
         type: AverageValue
     type: Pods
+  - pods:
+      metric:
+        name: nginx_http_requests_total
+      target:
+        averageValue: "4"
+        type: AverageValue
+    type: Pods

This gets even worse if you have two metrics, now an empty metric is added by the Kubernetes API:

$ kubectl diff --server-side -f hpa.yaml
+  - pods:
+      metric:
+        name: nginx_http_requests_total
+      target:
+        averageValue: "4"
+        type: AverageValue
+    type: Pods
+  - type: ""

Ref: fluxcd/kustomize-controller#494

[darkowlzz]

LGTM!
Thanks.