Aligning testing process with release process

[chrislovecnm]

We are not using the same scripts for the release that we are using for testing. I just discovered a bug in the Makefile where we have cockroachdb-operator instead of cockroach-operator. I recommend that we align the container release and the OpenShift packing release process with how I am testing the OpenShift operator packaging.

@rail / @udnay thoughts?

[rail]

We definitely should use the same process. I'd even say, CI should build images ready to be released without any modifications. In other words we would "promote" CI artifacts as release artifacts. This way we would release the bits waht we tested.

The release process tries to reuse as many Makefile targets as possible. ATM we have 3 distinct steps:

1. Publish the operator docker image to dockerhub. It uses a single make call
2. Publish the operator image to RedHat. This one does not use any make targets, instead it simply pulls the image from dockerhub and pushes it to Redhat.
3. Publish the bundle to RedHat. This step doesn't use any make targets, instead, it uses docker build/tag/push. From the comments in the script it sounds like bazel doesn't support the FROM: scratch type docker images (yet?).