Polish (containerized) roxie UX for OpenShift

[mclasmeier]

Addresses #39

User-facing:

• Don't require image pull credentials when deploying to Infra OpenShift (because they are already on the cluster). This applies to the non-containerized roxie as well.
• Emit user-friendly messages in case something is wrong with the provided kubeconfig (e.g. in case the :U suffix for ownership changing is not provided). This change includes error messages specifically for the containerized workflow.

Most of the changes are in the internal/env package, specifically around the environment detection (cluster type detection). This has been reorganized to now require explicit initialization with a logger instead of implicit initialization on first use without any logger.

This PR allows simplified one-command deployments to an Infra OpenShift clusters, e.g.:

$ podman run --rm -it --privileged -v $KUBECONFIG:/kubeconfig:U \
  quay.io/rhacs-eng/roxie:0.1.0-20-g84b295b deploy -t 4.11.x-39-g9aeddf5917 --olm --resources=auto

Note that we are not passing REGISTRY_USERNAME and REGISTRY_PASSWORD anymore.

[mclasmeier]

I got @kcarmichael08 to test out this version using the containerized workflow against an OpenShift cluster. Worked. 🎉

[vladbologa]

LGTM with some nits

[mclasmeier]

@vladbologa Addressed two comments, one is unclear to me. I had to touch one existing existing integration test, which would panic with the new behaviour when executed without valid kubectl context.