Re-evaluate what jobs we want to run for CI changes

We have been talking about this, but I couldn't find an issue for it, and it's brought up most recently in !120566 (comment 1402290777)

Let's not try to run all kinds of jobs when we're making CI changes. Yes conceptually CI changes affect everything, but it's not going to scale if we run everything for even the smallest change. We need to be more precise.

For example, for db-patterns, which is looking at .gitlab/ci/**/* now, perhaps we can just look at .gitlab/ci/rails.gitlab-ci.yml because most database related tests are defined over there.

More importantly, validating the CI config should not be achieved by running the full test suites. We have quite some issues talking about how to validate CI changes:

• #34040 (closed)
• gitlab-org/quality/engineering-productivity/team#74 (closed)
• And more

If we want to run specific jobs to see if they're running correctly, we can perhaps use labels like pipeline:run-single-db and pipeline:run-praefect-with-db for the respective jobs we're looking for. This way even though we cannot test it automatically when we're making CI changes, we can at least manually toggle them to see.

We might not want to have a label for each respective jobs, but something like that to manually toggle jobs can be useful and a human should know what they're doing.