Change RHCOS booting for AWS to work like GCP/Azure

[cgwalters]

For GCP and Azure and other clouds, all we upload is a blob to storage. The installer creates "bootable images" from that.

But for historical reasons, the RHCOS pipeline creates AMIs directly - the installer just uses them for the bootstrap node, creating a separate encrypted AMI for the actual cluster.

I think we should do the encryption bit before launching the bootstrap, and then RHCOS can stop publishing AMIs, and things will work more consistently across clouds.

[pierreprinetti]

/label platform/aws

[cgwalters]

Also xref this comment: coreos/mantle#1168 (comment)

Basically...today the *COS pipelines only sanity check in AWS - for RHCOS it's convenient because we have AMIs, but as soon as we don't, this section will break:
https://docs.openshift.com/container-platform/4.3/installing/installing_aws/installing-aws-user-infra.html#installation-aws-user-infra-rhcos-ami_installing-aws-user-infra

Basically as the issue notes we want openshift-install create-bootable or so - until we have that, doing UPI installs on !AWS is a lot more painful and requires users to dive into the rhcos.json.

[cgwalters]

Although as of lately, #3293 landed which pushes things back towards using AMIs. So...I guess we should close this.