This was causing too many false positives which are not representative
of code that is actually running in production. It is very common in
tests to build a tree of related objects in FactoryBot and save them all
in one go with `save!` (this is the default behaviour in FactoryBot for
association building). When this is done it creates an outer transaction
and saves all the models within the context of this transaction. If 2 of
the models belong in a different `gitlab_schema` (database) then this
triggers our cross-database modification detection logic. While there is
a chance that this could be indicative of problems in our model code
(eg. `before_save` and so on) it is more often than not just a problem
with how our factories are building a relation tree.

A simple example is included as an RSpec test here where we create a
`ci_runner` and `project` (which in turn creates a `user`) all in one
go. This is convenient in tests but it is never going to happen in real
production code as there is no way to create a project at the same time
as creating a runner (let alone user). You would always have a project
and a user and then later create the runner for that project.

Our PreventCrossDatabaseModification detection logic is designed to find
places in our code where we are writing to 2 different databases in the
context of a transaction of one of those databases. This was introduced
because we are decomposing our database into separate databases.
Specifically we are moving all CI tables to a separate CI database. It's
important that we find and fix any places in our code where we are
opening a transaction and expecting all transaction semantics (ie.
rollbacks) to apply to all database queries within that context. When
you have 2 databases there may be places where we are only rolling back
some of the queries in that transaction which could lead to data
inconsistency bugs. As such we need to detect and restructure such
transactions so that each case is properly handled.

Revert "Merge branch 'yorick/clean-up-merge-request-diff-commits-migration' into 'master'"

See merge request gitlab-org/gitlab!73035

Alphabetized artifacts reports keywords

See merge request gitlab-org/gitlab!73022

Add a list of how account creation

See merge request gitlab-org/gitlab!73014

Fix `online_garbage_collection_spec.rb` QA failure

See merge request gitlab-org/gitlab!72888

Shrink cross-db spec allowlist

See merge request gitlab-org/gitlab!73031

Add mocha to javascript examples

See merge request gitlab-org/gitlab!72963

This reverts merge request !72219

Polish wording of Pseudonymizer UI

See merge request gitlab-org/gitlab!72613

Feature page styling

See merge request gitlab-org/gitlab!72976

Final shrinkage of spec allowlist, the rest arises from specs somehow,
with the exception of package specs which is known.

Reshape the text - what little there is - for the Pseudonymizer
option in the UI. Add a 'learn more' link for users who don't know
what this item is. Docs work is also needed, but that can happen
in the documentation.

Remove duplicate text from create group description

See merge request gitlab-org/gitlab!72954

Show linked pipelines mini graph in the pipeline editor

See merge request gitlab-org/gitlab!72967

Do not promise additional language support for Dependency Scanning

See merge request gitlab-org/gitlab!72966

ci: Don't install gems in frontend jobs

See merge request gitlab-org/gitlab!72992

Related to: https://gitlab.com/groups/gitlab-org/-/epics/6941

Close the registration dropdown after a token reset

See merge request gitlab-org/gitlab!73009

Redirect Geo HTTP(s) pulls to different path with missing repo

See merge request gitlab-org/gitlab!72881

Optimize JIRA ref lookup

See merge request gitlab-org/gitlab!72739

Remove retarget_merge_requests feature flag

See merge request gitlab-org/gitlab!72782

In a project with JIRA activated, `ProcessCommitWorker` attempts to add
a comment to a JIRA issue if that issue is mentioned in a
commit. However, the JIRA integration would attempt to retrieve all tags
and branches and pick the first matching ref given a commit OID.

The problem with that approach is that after each push, the list of all
branches and tags are expired and could take a while to gather. Since
multiple `ProcessCommitWorker` jobs can be running at the same time,
this can lead to high I/O on Gitaly nodes since multiple
`ProcessCommitWorker` jobs can run at the same time.

We observe that we don't really need to build the entire ref list; we
can just ask Gitaly for a single matching ref for the given OID with the
newly-created `FindRefsByOID` RPC introduced in
https://gitlab.com/gitlab-org/gitaly/-/merge_requests/3947.

Relates to https://gitlab.com/gitlab-org/gitlab/-/issues/343035

Changelog: performance

Doc Consistency: user/admin_area - source code

See merge request gitlab-org/gitlab!72058

Doc Consistency: user/discssions - code review

See merge request gitlab-org/gitlab!72936

Update one filename and one Screenshot

See merge request gitlab-org/gitlab!72973

This file has a dot instead of an underscore, and it's breaking
links. Renaming the file.

Enable `assumeImmutableResults` in `app/assets/javascripts/terraform/index.js`

See merge request gitlab-org/gitlab!68571

Docs: Fix names of environment variables

See merge request gitlab-org/gitlab!72307

Persist descriptive fields in `Mutations::Vulnerabilities::Create`

See merge request gitlab-org/gitlab!72883

Fix Amazon RDS inconsistency

See merge request gitlab-org/gitlab!71094

Fix error 500 loading branch with UTF-8 characters with performance bar

See merge request gitlab-org/gitlab!72925

Annotate DAST models strings for localization

See merge request gitlab-org/gitlab!72799

Changelog: added
MR:
EE: true

Persist card network

See merge request gitlab-org/gitlab!72272