Feature/processor output individual records for gratia, add CPU usage and memory metrics

[mwestphall]

• Add a CPU usage metric based on container_cpu_usage_seconds_total.
  • NB: on old clusters using cgroups v1 this double counts results
• Add a new memory calculation using the kube_pod_container_resource_requests{resource="memory"} metric, similar to cores.
• Update KAPEL.py to optionally output individual APEL job records if gratia is being used for output. This aligns better with the expected GRACC record format
• Add a simple Dockerfile for building the processor InitContainer

@rptaylor The commit history is a bit messy for this change since I separated it out from the helm chart changes in-post, let me know if you'd like it condensed down with a rebase.

[rptaylor]

LGTM

[rptaylor]

Actually for the record I believe the cgroup-related differences we saw are due to using EL8 where the default is still cgroup v1, not because of the k8s version per se.