Compare the outcome of all API test case scenarios with and without the metadata database

Problem

During the development of the metadata database, we found some situations where the database-backed API operations were not following the behavior of the filesystem-backed API, such as #173 (closed) and #152 (closed).

Solution

We should go over every possible API test case scenario (i.e., every possible combination of operation and corresponding result) and compare the outcome between a registry without the database enabled and another with the database enabled. We should look for possible holes in test coverage, paying particular attention to test cases around authorization (e.g., cross repository blob mounting).

This work doesn't include fixing possible issues, it's only meant for testing and reporting.