Revert "Revert "Merge branch 'support-go-subpackages' into 'master'""

This reverts commit 5a586f36

See merge request !3194

This reverts commit 5a586f36

Remove unnecessary go-get test (superseded by middleware)



See merge request !3193

This reverts merge request !3191

Support Golang subpackage fetching

Closes #13805

See merge request !3191

Removed User#project_relations

This is to make things more compatible with EE so merging is easier.

cc @dzaporozhets 

See merge request !3192

GitLab EE adds an extra relation that selects a "project_id" column
instead of an "id" column, making it very hard for this method to be
re-used in EE. Since using User#authorized_groups in
ProjectsFinder#all_groups apparently has no performance impact we can
just use it and keep everything compatible with EE.

Filter import_url params because they may contain auth information. Fixes #14199

cc @dzaporozhets @DouweM tagging you since this is security related.

See merge request !3174

Fix importing PR's from GitHub when the source repo was removed

Closes #13847 
Closes gitlab-com/support-forum#584

See merge request !3172

Disallow blank (non-null) values for a Note's `line_code` attribute

It's unclear how these blank values got added, but GitLab.com had a few:

```
irb(main):002:0> Note.where("line_code IS NOT NULL AND line_code = ''").count
=> 439
```

We've added a migration to convert any existing records to use a NULL
value when blank, and updated Note to set blank values to nil before
validation.

See merge request !3118

Ignore .byebug_history

This allows me to use `git add .` without problems and retain all my history of `byebug`

What do you think @DouweM?

See merge request !3145

fix token issue - timing attack

Updates token comparisons to use a secure version instead of `==`

Fixes https://gitlab.com/gitlab-org/gitlab-ce/issues/13617

See merge request !3062

# Conflicts:
#	app/helpers/application_helper.rb

Closes #13805

Fix responsive bug top navigation

Hi,

I have found and fixed a minor layout bug (responsive)

This is the situation before the fix:
![before](/uploads/55bf61a64fb49a763c5f8381bf46d390/before.png)

And this is the situation after the fix:
![after](/uploads/33ba091156d404e52c41f134121c59f0/after.png)

See merge request !3190

It's unclear how these blank values got added, but GitLab.com had a few:

```
irb(main):002:0> Note.where("line_code IS NOT NULL AND line_code = ''").count
=> 439
```

We've added a migration to convert any existing records to use a NULL
value when blank, and updated Note to set blank values to nil before
validation.

Use a custom Devise failure app to handle unauthenticated .zip requests

Closes https://gitlab.com/gitlab-org/gitlab-ce/issues/12944

See merge request !2828

Refactor searching and use PostgreSQL trigram indexes for significantly
improved performance

Related issue: https://gitlab.com/gitlab-org/gitlab-ce/issues/13743.

Also fixes #12410

See merge request !2987

This explains the user what they need to run and where to go in case
they want to learn more about "CREATE EXTENSION".

The OR condition for source_project_id/target_project_id leads to a
query plan that performs rather poorly on PostgreSQL due to the use of
sub-queries. Because Rails offers no easy alternative for this
particular problem we're forced to using a UNION for both conditions.
The resulting query performs much faster than just using an OR.

This ensures searching namespaces works exactly the same as searching
for any other resource.

We can just use "arel_table" in these cases instead of
"SomeClass.arel_table".

We don't need the extra layer of nesting of UNION queries here (as
User#authorized_projects already returns a UNION'd query).

There's no need to order queries used as sub-queries and doing so can
add potential overhead.

This ensures some other methods such as the "issues" method still work.

This spec was still passing an ID to the #initialize method instead of
a Project instance.

This code is mostly a copy-paste from existing pull requests so there's
no point in running Rubocop on it.

This ensures that options such as `using: :gin` and PostgreSQL operator
classes are ignored when loading a schema into a MySQL database.

This also includes e.g. the appearances table which apparently wasn't
already included in the schema.

This is needed to support creating/dumping/loading indexes that use the
gin_trgm_ops operator class on PostgreSQL. These changes are taken from
Rails pull request https://github.com/rails/rails/pull/19090.