Kubernetes memory limits ought to be greater then requests

[adamnovak]

The Kubernetes way of doing things is to give a bit of slack to processes and set limits that are greater than the requests, to avoid killing jobs unnecessarily, while making efficient use of resources, by allowing a slight overcommit. We should maybe do that when we issue jobs.

Also, I have a run I want to resume where my memory limits are too low and the only way I can get my jobs to run is if Kubernetes gives them a bit more memory.

┆Issue is synchronized with this Jira Task
┆Issue Number: TOIL-459