Extend instructions on adding a Profile by adding LimitRange

[gkcalat]

          @jsitu777 thank you for reporting this.

Profile controls user workloads to be within the quota. So, if you have a hard quota on CPU and memory it will reject all workloads that do no satisfy it. Unfortunately, it seems that Containers without resources will also be rejected.

Adding resources to the Argo's workflow-controller-configmap results in all containers being overwritten and having the same resources, which is probably a wrong way to fix the issue described above.

I feel we may have better solutions:

1. Allowing default values in Profile, which will be used to fill resources for containers that do not have them. This needs to be done in Profile component. /cc @kimwnasptd
2. Advising users to create a LimitRange whenever a Profile gets created. We need to update the documentation.

Example LimitRange:

apiVersion: v1
kind: LimitRange
metadata:
  name: <user>-limit-range
spec:
  limits:
  - default:
      cpu: 0.5
      memory: 512Mi
    defaultRequest:
      cpu: 0.01
      memory: 32Mi
    type: Container

Which can be applied a user namespace:

kubectl apply -f <yaml> -n <user>

/cc @chensun @Linchin

Originally posted by @gkcalat in kubeflow/pipelines#9089 (comment)