Add nightly comparison testing with key images

It would be ideal to collect a key set of docker images and run comparison testing on the results against a known good result nightly. Today we do this in the acceptance pipeline, however, the target JSONs are checked directly into the repo, which will not work with larger results without preprocessing to prune the results to only what needs to be compared (type-name-version).

Start by reviewing the acceptance pipeline and the checked-in target JSONs mentioned in the issue. Define how the key Docker image set and nightly comparison should work, including preprocessing results to type-name-version; done means larger results can be compared against known-good outputs without storing unpruned targets.

Written by the indexing model from the issue text.