Improve EBS CSI driver release process

[torredil]

What happened?

• The existing release process involves having to manually build the multi-arch CSI driver container image when promoting to a private ECR staging registry which is both inefficient and prone to human error. Additionally, the image is manually built again at a later step when pushing to public ECR to preserve tags for platform and architecture specific images, Stop tagging OS/architecture specific images in public ECR #1315.

aws-ebs-csi-driver/Makefile

Line 85 in 0c045fd

# "Currently, multi-platform images cannot be exported with the docker export type. The most common usecase for multi-platform images is to directly push to a registry (see registry)."

What you expected to happen?

• Ideally, the steps to build and push CSI driver container images to the private ECR staging registry should be automated. Furthermore, the container images should be copied over from the staging registry to the public registry. This will result in higher quality deployments with less chance for error, also fixes Stop tagging OS/architecture specific images in public ECR #1315.

• The diagram below outlines the proposed solution.

Anything else we need to know?:

• The internal CodePipeline is triggered by Amazon EventBridge once new images are pushed to the private ECR staging registry.

cc @jaypipes