cloud-init runcmd keeps running even when a command fails

[cmd-ntrf]

All commands in MC's cloud-init are essential to the puppet configuration that follows on next boot.

While these commands work in normal conditions, sometime cloud-init is executed in abnormal conditions, like without an internet connection. This results in some of the commands failing and puppet being unable to run properly at next boot.

runcmd should probably stop on first error and a re-run of cloud-init should be triggered.

[cmd-ntrf]

Examples of error handling in a bash script:
https://intoli.com/blog/exit-on-errors-in-bash-scripts/

[cmd-ntrf]

After 3 years of reflection, we figure that we should let runcmd run all its command, but log what command failed and prevent the instance from rebooting so the operator may fix the issues manually and reboot.

This is what is implemented in #403