Allow fast task startups / kills

[adersberger]

Currently youve to wait between button activations/deactivations until a deployment is fishished. Otherwise Kubepad can't deploy because the app is locked. We should implement a grace period where Kubepad waits for a few 100rds of ms for a new button hit and then batch them together.

[voidc]

Improved in 37343f8

This works because when scaling up, a new Deployment is created for each new instance (e.g. 3->4, 4->5, 5->6). The force option allows new Deployments to cancel current deployments. This means that only the last of multiple deployment requests in a short period of time will have an effect.

[lreimer]

Does this also work when scaling down?