Backup pods should make use of the fdbmonitor/fdb-kubernetes-monitor to start processes

[johscheuer]

What would you like to be added/changed?

Currently the operator creates a deployment to manage the backup agents as a Kubernetes Deployment. In this deployment the backup_agent process is directly started. This has a few drawbacks, e.g. the backup_agent will run as PID 1 and not listen on signals, which makes it impossible to kill the process (and the signal for a graceful shutdown will be ignored). To prevent this we should use the monitor process to start the backup_agent, otherwise the handling of the deployment strategy should be through the Kubernetes Deployment controller.