feat(409): opt-out of auto-scaling per server (k8s)

[moritzrinow]

This PR addresses #409 and adds the possibility to selectively disable auto-scaling via annotations on Kubernetes.

Annotations:

• mc-router.itzg.me/autoScaleUp for controlling scale-up
• mc-router.itzg.me/autoScaleDown for controller scale-down

As mentioned, it is currently only possible to opt-out of auto-scaling, as the configuration flags --auto-scale-up and --auto-scale-down decide whether or not the actual watcher service is being instantiated or not. However, the piece of logic I added should also work for opt-in if it is being called.

I chose the annotations to work on Service rather than StatefulSet because the service seems like the main abstraction for the server configuration and the stateful-set is only the backend.

@itzg unfortunately I had to migrate the version of package golang.ngrok.com/ngrok down to v1.12.1 because I was not able to build for v1.13.0. It was complaining that this rev does not exist anymore. I checked the Github repo and it does indeed not exist.

[moritzrinow]

Technically with this, you can somewhat opt-in to auto-scaling if one of the two (scale-up or scale-down) is enabled via flag.
Example:
One can use the annotation mc-router.itzg.me/autoScaleUp=true to opt-in to auto scale-up if the flag --auto-scale-down was provided, since the watcher is active then.

[itzg]

Thanks!

[itzg]

Released in https://github.com/itzg/mc-router/releases/tag/v1.36.0