CNI workload: Change the probe from readinessProbe to startupProbe.

This patch changes the probe in curl-deployment.yml from
readinessProbe to startupProbe for the following reasons:

 - readinessProbe is periodic and with default period interval of 1
   second, it significantly loads kubelet effecting the pod
   readiness as the scale increases.

 - startupProbe is enough to check if CNI has done its job of wiring
   the networking of the pods.

 - Below are the tests on AWS 120 node cluster with 4000 iterations
   of kubelet-density-cni

   With readinessProbe - ContainerReady
   -----------------------------------
    "P99": 80928,
    "P95": 66161,
    "P50": 9526,
    "max": 671045,
    "avg": 19321,

   With startupProbe - ContainerReady
   -----------------------------------
    "P99": 11872,
    "P95": 9609,
    "P50": 4862,
    "max": 18487,
    "avg": 5329,

Signed-off-by: Numan Siddique <nusiddiq@redhat.com>

examples/workloads/kubelet-density-cni-networkpolicy/templates/curl-deployment.yml

@@ -22,13 +22,15 @@ spec:
         imagePullPolicy: IfNotPresent
         securityContext:
           privileged: false
-        readinessProbe:
+        startupProbe:
           exec:
             command: 
               - "/bin/sh"
               - "-c"
               - "curl ${WEBSERVER_HOSTNAME}:${WEBSERVER_PORT}"
           periodSeconds: 1
+          timeoutSeconds: 60
+          failureThreshold: 600
       restartPolicy: Always
   replicas: 1
   selector:

examples/workloads/kubelet-density-cni/templates/curl-deployment.yml

@@ -22,13 +22,15 @@ spec:
         imagePullPolicy: IfNotPresent
         securityContext:
           privileged: false
-        readinessProbe:
+        startupProbe:
           exec:
             command: 
               - "/bin/sh"
               - "-c"
               - "curl ${WEBSERVER_HOSTNAME}:${WEBSERVER_PORT}"
           periodSeconds: 1
+          timeoutSeconds: 60
+          failureThreshold: 600
       restartPolicy: Always
   replicas: 1
   selector: