fix: lower gpu node cpu-request to allow scheduling

[blairdrummond]

My colleague's GPU image was not able to be scheduled, it appears because of the CPU limits. After doing a few kubectl edits on his image, adjusting the cpu and memory, I found that reducing the CPU without touching the memory was enough to get the image scheduled.

Closes #58

[blairdrummond]

This is an out-of-the-box GPU image

blair@system76-pc ~/jupyter-apis (lower-gpu-cpu-request)$ k describe pod -n blair-drummond x-0 | tail
                 node.kubernetes.io/not-ready:NoExecute for 300s
                 node.kubernetes.io/unreachable:NoExecute for 300s
Events:
  Type     Reason             Age                   From                Message
  ----     ------             ----                  ----                -------
  Warning  FailedScheduling   <unknown>             default-scheduler   0/41 nodes are available: 2 Insufficient pods, 35 Insufficient cpu, 4 Insufficient memory.
  Warning  FailedScheduling   <unknown>             default-scheduler   0/41 nodes are available: 2 Insufficient pods, 35 Insufficient cpu, 4 Insufficient memory.
  Normal   NotTriggerScaleUp  12m (x3 over 14m)     cluster-autoscaler  pod didn't trigger scale-up (it wouldn't fit if a new node is added): 1 Insufficient nvidia.com/gpu, 1 Insufficient cpu, 1 Insufficient memory
  Normal   NotTriggerScaleUp  10m (x5 over 14m)     cluster-autoscaler  pod didn't trigger scale-up (it wouldn't fit if a new node is added): 1 Insufficient memory, 1 Insufficient nvidia.com/gpu, 1 Insufficient cpu
  Normal   NotTriggerScaleUp  4m33s (x44 over 14m)  cluster-autoscaler  pod didn't trigger scale-up (it wouldn't fit if a new node is added): 1 Insufficient cpu, 1 Insufficient memory, 1 Insufficient nvidia.com/gpu

[blairdrummond]

@brendangadd any objections to merging?

[blairdrummond]

@brendangadd I think this is good, but don't know anything about how this gets deployed

[ca-scribner]

per @wg102, this gets deployed via the manifests. So merging this in is safe

[ca-scribner]

PR to deploy here: StatCan/aaw-kubeflow-manifests#86