These can be a `BatchLoader` which is proxying a nil, while not being concrete
nils themselves.

Closes #44198

and fixes the project description not being imported

The specs are based on a schema version that doesn't know about
`internal_ids` table. However, the actual code being execute relies on
it.

Attempted commits were failing due to the pre-recieve LfsIntegrity
check. This is skipped during tests making this failure silent.

Closes #43355

This is most a backport of
https://gitlab.com/gitlab-org/gitlab-ee/merge_requests/4987/diffs but includes
other columns that use column_exists? in a way that may cause unnecessary
schema loads.

This should use stub_omniauth_setting, not do it manually.

Carried over from https://gitlab.com/gitlab-org/gitlab-ce/merge_requests/8857

This may lead to some being evicted and having to be cached again, but many MRs
get closed or updated in that time anyway.

Previously, we kept them all in the cache. We don't need the highlight results
for older diffs - if someone does view that (which is rare), we can do the
highlighting on the fly.

Updates Files::MultiService for the commits API which is in turn used by the
multi-file upload web UI

Ensures that files which should be in LFS are transformed into LFS pointers
Uses Lfs::Transformer which then links LfsObjectProjects on success

When importing from a GitLab archive, an admin can carry over the
assignations. Other users can not.

When a regular user is importing an issue with multiple assignees, the
assignee is replaced with the current user, meaning we would try to
insert current user as an assignee multiple times.

By filtering the array before storing it, the import becomes more
robust.