Restore role follower label to eliminate failures in kubernetes-conjur-demo

[diverdane]

The conjurdemos/kubernetes-conjur-demo repo CI is currently failing on that repo's master. Here's a recent failure: https://jenkins.conjur.net/job/conjurdemos--kubernetes-conjur-demo/job/master/373/ The console output from one test case shows this error:

09:54:10  + ./4_store_conjur_cert.sh
09:54:10  ++++++++++++++++++++++++++++++++++++++
09:54:10  
09:54:10  Storing Conjur cert for test app configuration.
09:54:10  
09:54:10  ++++++++++++++++++++++++++++++++++++++
09:54:10  Retrieving Conjur certificate.
09:54:10  No resources found.
09:54:11  Value added
09:54:11  Value added
09:54:11  No resources found.
09:54:11  error: expected 'exec POD_NAME COMMAND [ARG1] [ARG2] ... [ARGN]'.
09:54:11  POD_NAME and COMMAND are required arguments for the exec command
09:54:11  See 'oc exec -h' for help and examples.

The problem stems from this section of the deploy script 4_store_conjur_cert.sh:

if $cli get pods --selector role=follower --no-headers; then
  follower_pod_name=$($cli get pods --selector role=follower --no-headers | awk '{ print $1 }' | head -1)
  ssl_cert=$($cli exec $follower_pod_name -- cat /opt/conjur/etc/ssl/conjur.pem)
else
  echo "Regular follower not found. Trying to assume a decomposed follower..."
  follower_pod_name=$($cli get pods --selector role=decomposed-follower --no-headers | awk '{ print $1 }' | head -1)
  ssl_cert=$($cli exec -c "nginx" $follower_pod_name -- cat /opt/conjur/etc/ssl/cert/tls.crt)
fi

Apparently, we need the role: follower (and possibly the name: conjur-follower) pod template labels added back in for the demo scripts to be able to find the follower based on the role: follower label. It should look like this:

 template:
    metadata:
      labels:
        app: conjur-follower
        name: conjur-follower
        role: follower