Generating EOL annotation data for images is slow. Currently the command takes in a list of image digests, and then for each digest:

• Uses the oras CLI to check if the image already has a lifecycle annotation.
• Uses the oras CLI to add the lifecycle annotation if the image doesn't already have one.

Even though this operation is parallelized (with Parallel.ForEach), it's really slow for large quantities of images. This makes our publish jobs and ACR cleanup pipelines take a long time.

Related:

• cleanup-acr-images pipeline is timing out due to running for over 1 hour #1827