Better support for executing K8S jobs

[NicolasRouquette]

Since many things can go wrong when executing a K8S job, it is important to monitor a job for progress of all of its containers, including init containers. Finally, it is important to clean up a job when it's executed; however, since deleting kubernetes resources takes time, one needs to monitor the deletion of a job as well.

Another consideration is that skuber hides the host connection pools it creates internally for long-duration operations (e.g., watchContinuously). Although the Akka API underneath will reuse the pool, this results in unnecessary garbage objects and skuber neither shutdowns these pools nor exposes them for the user to shutdown.

[NicolasRouquette]

This issue was incorrectly addressed in this PR: #274
Made a new PR to explicitly target this issue.

[opensorceror]

+1

[doriordan]

As discussed with Nicolas, I am working on incorporating a means to do this into a stream of work I am currently progressing on enhanced list/watch based features, will update this issue in due course