Previous versions did not properly set the IAT leeway properly, causing
many Geo API requests to be marked invalid. The problem was that
the ruby-jwt `master` used `iat_leeway` as the parameter, but v1.5.6
actually used `leeway`.

Use curl --remote-name instead of curl -O to fix broken docs lint

See merge request !2209

CE to EE: refactor projects finder init collection ee

See merge request !2199

EE version: Clean up the Project model to reduce conflicts

See merge request !2056

Moves remote mirror, import and fork workers to their own queue

See merge request !2177

Add Rubocop Style/DotPosition Cop

See merge request !2201

The spec filename does not match the module/class name. But ignore it, because
the `ee` subdir is added in between to avoid conflicts when merging from CE.

Fix typo with custom OpenSSH build instructions

See merge request !2202

Namespace license checks for exporting issues (EES)

Closes #2584

See merge request !2164

Fix transient remove issue from board and see milestone removed spec

Closes #2660

See merge request !2139

[ci skip]

Fix Vue warnings in the MR Widget (EE)

See merge request !2192

Fix Vue warnings about string idKey in Related Issues

See merge request !2190

Add instructions for setting up a custom version of OpenSSH for Geo

See merge request !2191

CE and EE implement this differently, so abstract it out of VisibilityLevel.

To limit the number of CE merge conflicts, move the auditor role code outside
the original User model and into the EE-specific mixin.

The permitted visibility levels for an auditor user should also include private,
because it has read-only access to every project/group on the GitLab instance.

This optimises how GroupProjectsFinder builds it collection, producing
simpler and faster queries in the process. It also cleans up the code a
bit to make it easier to understand.

This method used to use a UNION, which would lead to it performing the
same query twice; producing less than ideal performance. Further, in
certain cases ActiveRecord could get confused and mess up the variable
bindings, though it's not clear how/why exactly this happens.

Fortunately we can work around all of this by building some of the WHERE
conditions manually, allowing us to use a simple OR statement to get all
the data we want without any of the above problems.

This changes ProjectsFinder#init_collection so it no longer relies on a
UNION. For example, to get starred projects of a user we used to run:

    SELECT projects.*
    FROM projects
    WHERE projects.pending_delete = 'f'
    AND (
        projects.id IN (
            SELECT projects.id
            FROM projects
            INNER JOIN users_star_projects
                ON users_star_projects.project_id = projects.id
            INNER JOIN project_authorizations
                ON projects.id = project_authorizations.project_id
            WHERE projects.pending_delete = 'f'
            AND project_authorizations.user_id = 1
            AND users_star_projects.user_id = 1

            UNION

            SELECT projects.id
            FROM projects
            INNER JOIN users_star_projects
                ON users_star_projects.project_id = projects.id
            WHERE projects.visibility_level IN (20, 10)
            AND users_star_projects.user_id = 1
        )
    )
    ORDER BY projects.id DESC;

With these changes the above query is turned into the following instead:

    SELECT projects.*
    FROM projects
    INNER JOIN users_star_projects
        ON users_star_projects.project_id = projects.id
    WHERE projects.pending_delete = 'f'
    AND (
        EXISTS (
            SELECT 1
            FROM project_authorizations
            WHERE project_authorizations.user_id = 1
            AND (project_id = projects.id)
        )
        OR projects.visibility_level IN (20,10)
    )
    AND users_star_projects.user_id = 1
    ORDER BY projects.id DESC;

This query in turn produces a better execution plan and takes less time,
though the difference is only a few milliseconds (this however depends
on the amount of data involved and additional conditions that may be
added).

This ensures that Project only includes a single prepend for
EE::Project, which in turn takes care of the rest.

* master: (26 commits)
  Document the manual steps needed to use repmgr
  Update CHANGELOG.md for 9.2.7
  Update CHANGELOG-EE.md for 9.2.7-ee
  Change GitLab version for object storage
  Speed up counting approvers when some are specified
  Fix bad connector - verify if both new and resolved issues are present before adding the connector.
  Re-enable autocomplete for milestones, tags, releases, and wiki
  Clarify that only when using CI_JOB_TOKEN multi-project pipelines exist
  Add link to issue for MySQL/subgroups drop support
  Fix duplicate testing of vm.triggered.
  Fix avatar images in pipeline and approval emails
  Fix some failing specs on master
  Change a few missing `login_as` to `gitlab_sign_in`
  Shush eslint.
  Add specs for coercing triggeredBy into a collection.
  Fix up specs that depend on triggered_by.
  Fix up propsData refs to triggered_by.
  Shush eslint.
  Update expected format for linked_pipelines_mack_data.
  Coerce triggered_by response into collection.
  ...

Conflicts:
	spec/helpers/application_helper_spec.rb

This moves all EE specific code out of the Project model into separate
modules. State machine logic is injected using the
EE::Project::ImportStatusStateMachine module. By moving all this code
out of the Project model we reduce the number of possible merge
conflicts.

Speed up counting approvers when some are specified

See merge request !2196

Change GitLab version for object storage

See merge request !2197

Document the manual steps needed to use repmgr

See merge request !2172

[ci skip]

[ci skip]

[ci skip]

When some approvers are specified, we still need to check the potential set of
approvers, including people with developer access and up. Doing that with an OR
took a long time (over 500ms) on GitLab.com, even when only one approver was
specified.

By doing a UNION instead, we get much faster results (less than 10ms). Also, as
we are using UNION and not UNION ALL, we don't need to exclude the approvers
specified from the query to get everyone with access.

Fix bad connector - verify if both new and resolved issues are present before adding the connector.

Closes #2676

See merge request !2187