CopyAcrImagesCommand should base its processing on image info, not manifest

Currently the logic in `CopyAcrImagesCommand` enumerates the contents of the manifest and correlates that data with the contents of the image info to determine what tags should be published. This can be problematic when bugs exist with that correlation (see https://github.com/dotnet/docker-tools/pull/710).

There is no reason for the logic to start with the manifest, however. This was done historically because the concept of image infos didn't even exist originally. The image info is really the source of truth as to what should be published. This won't completely eliminate the need for correlation between the manifest and image info due to the need to find syndicated repo data, but it can help in isolating it to `ImageInfoHelper`.

The basic idea here is that the current implementation has code which is unnecessary and can be refactored to simplify it. Simplifying the code means less potential for bugs.