Artifacts are in the middle of being migrated from ci_builds to
ci_job_artifacts. The expiration date is currently visible in both of
these tables and the test for whether an expired artifact is present
for a job is complex as it requires checking both the of the tables.

Add two new indexes, one on ci_builds.artifacts_expire_at and one on
ci_job_artifacts.expire_at to enable finding expired artifacts
efficiently.

And until the migration is finished, replace the SQL for finding
expired and non-expired artifacts with a hand-crafted UNION ALL based
query instead of using OR. This overcomes a database optimizer
limitation that prevents it from using these indexes.

When the migration is finished the next version should remove this
query and replace it with a much simpler query on just
ci_job_artifacts. See
https://gitlab.com/gitlab-org/gitlab-ce/issues/42561 for followup.

This should reduce the number of SQL queries and lookups needed to look up a
project of a build and pipeline and vice versa.

Before:

```
[1] pry(main)> Ci::Build.reflect_on_association(:project).has_inverse?
=> false
[2] pry(main)> Project.reflect_on_association(:builds).has_inverse?
=> false
[3] pry(main)> Ci::Pipeline.reflect_on_association(:project).has_inverse?
=> false
[4] pry(main)> Project.reflect_on_association(:pipelines).has_inverse?
=> :project
```

After:

```
[1] pry(main)> Ci::Build.reflect_on_association(:project).has_inverse?
=> :builds
[2] pry(main)> Project.reflect_on_association(:builds).has_inverse?
=> :project
[3] pry(main)> Ci::Pipeline.reflect_on_association(:project).has_inverse?
=> :pipelines
[4] pry(main)> Project.reflect_on_association(:pipelines).has_inverse?
=> :project
```

Use Class.new(StandardError) instead of custom extended error class. Bring back specified_dependencies?.

This reverts commit 714082e65304ae2ec5c5400c59a68ab63e724aa9.

Two things at ones, as there was no clean way to seperate the commit and
give me feedback from the tests.

But the model Artifact is now JobArtifact, and the table does not have a
type anymore, but the metadata is now its own model:
Ci::JobArtifactMetadata.

To allow jobs/builds to have multiple artifacts, and to start seperating
concerns from Ci::Build a new model is created: Ci::Artifact.

Changes include the updating of the ArtifactUploader to adapt to a
slightly different interface. The uploader expects to be initialized
with a `Ci::Build`.

Futher a migration with the minimal fields, the needed foreign keys and
an index.

Last, the way this works is by prepending a module to Ci::Build so we
can basically override behaviour but if needed use `super` to get the
original behaviour.

Compared to the merge_request_diff association:

1. It's simpler to query. The query uses a foreign key to the
   merge_request_diffs table, so no ordering is necessary.
2. It's faster for preloading. The merge_request_diff association has to load
   every diff for the MRs in the set, then discard all but the most recent for
   each. This association means that Rails can just query for N diffs from N
   MRs.
3. It's more complicated to update. This is a bidirectional foreign key, so we
   need to update two tables when adding a diff record. This also means we need
   to handle this as a special case when importing a GitLab project.

There is some juggling with this association in the merge request model:

* `MergeRequest#latest_merge_request_diff` is _always_ the latest diff.
* `MergeRequest#merge_request_diff` reuses
  `MergeRequest#latest_merge_request_diff` unless:
    * Arguments are passed. These are typically to force-reload the association.
    * It doesn't exist. That means we might be trying to implicitly create a
      diff. This only seems to happen in specs.
    * The association is already loaded. This is important for the reasons
      explained in the comment, which I'll reiterate here: if we a) load a
      non-latest diff, then b) get its `merge_request`, then c) get that MR's
      `merge_request_diff`, we should get the diff we loaded in c), even though
      that's not the latest diff.

Basically, `MergeRequest#merge_request_diff` is the latest diff in most cases,
but not quite all.

Adds a rubocop rule (with autocorrect) to ensure line break after guard clauses.