Introduce a churnOp to scheduler perf testing framework

- support two modes: recreate and create
- use DynmaicClient to create API objects

test/integration/scheduler_perf/config/churn/node-default.yaml

@@ -0,0 +1,11 @@
+apiVersion: v1
+kind: Node
+metadata:
+  generateName: node-churn-
+status:
+  capacity:
+    pods: "0"
+  conditions:
+    - status: "True"
+      type: Ready
+  phase: Running

test/integration/scheduler_perf/config/churn/pod-default.yaml

@@ -0,0 +1,8 @@
+apiVersion: v1
+kind: Pod
+metadata:
+  generateName: pod-churn-
+spec:
+  containers:
+  - image: k8s.gcr.io/pause:3.4.1
+    name: pause

test/integration/scheduler_perf/config/churn/service-default.yaml

@@ -0,0 +1,11 @@
+apiVersion: v1
+kind: Service
+metadata:
+  generateName: service-churn-
+spec:
+  selector:
+    app: foo
+  ports:
+  - protocol: TCP
+    port: 8080
+    targetPort: 8080

test/integration/scheduler_perf/config/performance-config.yaml

@@ -449,3 +449,29 @@
       initNodes: 5000
       initPods: 2000
       measurePods: 5000
+
+- name: SchedulingWithMixedChurn
+  workloadTemplate:
+  - opcode: createNodes
+    countParam: $initNodes
+  - opcode: churn
+    mode: recreate
+    number: 1
+    templatePaths:
+    - config/churn/node-default.yaml
+    - config/pod-high-priority-large-cpu.yaml
+    - config/churn/service-default.yaml
+    intervalMilliseconds: 1000
+  - opcode: createPods
+    countParam: $measurePods
+    podTemplatePath: config/pod-default.yaml
+    collectMetrics: true
+  workloads:
+  - name: 1000Nodes
+    params:
+      initNodes: 1000
+      measurePods: 1000
+  - name: 5000Nodes
+    params:
+      initNodes: 5000
+      measurePods: 2000

test/integration/scheduler_perf/config/pod-high-priority-large-cpu.yaml

@@ -0,0 +1,15 @@
+apiVersion: v1
+kind: Pod
+metadata:
+  generateName: pod-h-
+spec:
+  priority: 10
+  containers:
+  - image: k8s.gcr.io/pause:3.2
+    name: pause
+    ports:
+    - containerPort: 80
+    resources:
+      requests:
+        cpu: 9
+        memory: 500Mi