Add 'kubectl rabbitmq version' sub command

[ansd]

Relates #440.

Since we require the kubectl rabbitmq plugin to be installed via krew, the easiest way to output the plugin version is to ask krew. As explained in the commit message kubectl krew info rabbitmq will output the available - not installed - version. In contrast, kubectl krew list outputs the installed version.

As explained in kubectl krew list --help:

Remarks:
Redirecting the output of this command to a program or file will only print the names of the plugins installed

Therefore, we parse the rabbitmq plugin version from the krew list logs by increasing the verbosity to 4.

If the rabbitmq plugin is not installed via krew, the command will output an appropriate error message.

I wrote the tests such that they don't require krew to be installed. If krew isn't installed, we simply test for the error message.

Caveats:
If the rabbitmq plugin is installed via krew, but the script is not invoked via krew (e.g. cluster-operator/bin/kubectl-rabbitmq takes precedence over /usr/local/bin/kubectl-krew in PATH), the rabbitmq plugin version of krew will still be output. I think that's a minor issue for us developers, but not for end users.

Alternative approach:
Have some version variable hard coded in the bash script. Include a task in our Concourse release job which bumps that version and commits right before tagging. This approach seems to be more brittle because that task needs to commit directly to the main branch which introduces other difficulties (as have have experienced from our BOSH release pipelines).

[ChunyiLyu]

Nice!

[ablease]

Love it, thanks David!