Refactor K8sImpl structure to simplify the interaction with kubernetes

[Raffy23]

At the moment the interaction between the job-executor-service and kubernetes can be improved by refactoring the following things:

• The job-executor-service uses polling in the AwaitK8sJobDone function, we should be able to refactor this function to utilize the watch functionality from the kubernetes API and simplify a lot of logic when it comes to handling long running jobs.
• The K8sImpl structure can probably contain a lot of information which is passed around with the JobSettings structure, as these structures are initialized at the same time and the information in the JobSettings struct aren't changed.
• While it is necessary to pass the cloud event data around with the jsonEventData parameter, we should invest some time to provide a more structured way of accessing the contents of the cloud event.

With these changes we could not only improve the handling of long running / failed job states (e.g., #235 and #254), but also generalize the K8s interface such that it gets easier to mock functions for testing.