Postgres only needs to select a single column, so that can used as a
sub-query where `Milestone.upcoming_ids_by_projects` is actually used in
`IssuableFinder`.

MySQL needs to select the `due_date` column because it's used in the
`HAVING` clause, so it has to return an array of IDs.

Before: we took the next milestone due across all projects in the
search and found issues whose milestone title matched that
one. Problems:

1. The milestone could be closed.
2. Different projects have milestones with different schedules.
3. Different projects have milestones with different titles.
4. Different projects can have milestones with different schedules, but
   the _same_ title. That means we could show issues from a past
   milestone, or one that's far in the future.

After: gather the ID of the next milestone on each project we're looking
at, and find issues with those milestone IDs. Problems:

1. For a lot of projects, this can return a lot of IDs.
2. The SQL query has to be different between Postgres and MySQL, because
   MySQL is much more lenient with HAVING: as well as the columns
   appearing in GROUP BY or in aggregate clauses, MySQL allows them to
   appear in the SELECT list (un-aggregated).

Remove the annotate gem and delete old annotations

In 8278b763d96ef10c6494409b18b7eb541463af29 the default behaviour of annotation
has changes, which was causing a lot of noise in diffs. We decided in #17382
that it is better to get rid of the whole annotate gem, and instead let people
look at schema.rb for the columns in a table.

Fixes: #17382 

/cc @yorickpeterse 

See merge request !4099

In 8278b763d96ef10c6494409b18b7eb541463af29 the default behaviour of annotation
has changes, which was causing a lot of noise in diffs. We decided in #17382
that it is better to get rid of the whole annotate gem, and instead let people
look at schema.rb for the columns in a table.

Fixes: #17382

Fixes #3903