make systemd script bit more resilient

[natarajmb]

Some time packages fail to download due to network or certificate signing (slow processing on certs on RPI I think) and fails the package install. It would be nice to have the script to restart on such failures. It would need files copied into the root folder to be moved into a cleaner place to pick for the second run.

Apr 1 17:24:32 shimsha post_first_boot.sh[2757]: Err:6 https://packages.cloud.google.com/apt kubernetes-xenial Release Apr 1 17:24:32 shimsha post_first_boot.sh[2757]: Certificate verification failed: The certificate is NOT trusted. The certificate chain uses not yet valid certificate. Could not handshake: Error in the certificate verification. [IP: 216.58.210.238 443]

Though key was downloaded and added before above error like here
Apr 1 17:24:31 shimsha post_first_boot.sh[2567]: Warning: apt-key output should not be parsed (stdout is not a terminal) Apr 1 17:24:31 shimsha post_first_boot.sh[2567]: OK Apr 1 17:24:31 shimsha post_first_boot.sh[2753]: deb https://apt.kubernetes.io/ kubernetes-xenial main

[natarajmb]

Its sounds like my setup as most times this works and only few cases it fails as suggested here kubernetes/kubernetes#39592 hence making systemd rerun should be an ideal fix.