Refactor use of in memory registry in e2e test

[AustinAbro321]

Describe what should be investigated or refactored

#2869 changed the OCI e2e tests use a in memory database as we saw flakes during the EKS tests relating to connection resets on docker timeouts (#2837). There are a few things we may want to reconsider / refactor from this implementation however.

1. The distribution/distribution registry code doesn't give a way to close the server without sending a sigint to the server (as far as I can tell) . It's unclear when the server cleans up the resources.
2. All OCI artifacts are stored in memory, this shouldn't be a huge issue as the largest package currently in the test is 27mb, but this is something to be wary of if we test with larger packages.
3. We don't check the error from starting the registry server as it's in a go-routine. Potentially this could be resolved using error groups.