atomic 1.1.100 was yanked by its author.

This commit fixes the error:

```
  1) Admin::UsersController DELETE #user with projects deletes the user and their contributions when hard delete is specified
     Failure/Error: Users::DestroyService.new(current_user).execute(delete_user, options.symbolize_keys)

     NoMethodError:
       undefined method `symbolize_keys' for "{\"hard_delete\"=>\"true\"}":String
     # ./app/workers/delete_user_worker.rb:8:in `perform'
     # ./lib/gitlab/sidekiq_status/server_middleware.rb:5:in `call'
     # ./config/initializers/forbid_sidekiq_in_transactions.rb:35:in `block (2 levels) in <module:NoEnqueueingFromTransactions>'
     # ./app/models/user.rb:913:in `delete_async'
     # ./app/controllers/admin/users_controller.rb:148:in `destroy'
     # ./lib/gitlab/i18n.rb:50:in `with_locale'
     # ./lib/gitlab/i18n.rb:56:in `with_user_locale'
     # ./app/controllers/application_controller.rb:327:in `set_locale'
     # ./spec/controllers/admin/users_controller_spec.rb:28:in `block (3 levels) in <top (required)>'

Finished in 6.81 seconds (files took 13.9 seconds to load)
1 example, 1 failure
```

This commit does two things:
1. Synchronizes `Gemfile.rails5.lock` with the latest changes in `Gemfile`
2. Adds a monkey patch for active record to avoid type casting in uniqueness
validator.

In Rails 5.0 there was introduced a bug [1] that causes to problems like
this:

```
1) API::Users POST /user/:id/gpg_keys/:key_id/revoke when authenticated revokes existing key
Failure/Error: let(:gpg_key) { create(:gpg_key, user: user) }
TypeError:
  can't cast Hash
```

This bug was fixed in Rails 5.1 [2].

[1]: https://github.com/rails/rails/pull/23523/commits/811a4fa8eb6ceea841e61e8ac05747ffb69595ae
[2]: https://github.com/rails/rails/pull/24745/commits/aa062318c451512035c10898a1af95943b1a3803

1. Add support for `RAILS5=1|true` for the `bin/spinach` command.
2. Synchronize used spinach versions both for rails4 and rails5.

For rails5 it was accidently used spinach 0.10.1 instead of 0.8.10.
That brought some problems on running spinach tests.

Example of failure message:

NoMethodError: undefined method `line' for #<Spinach::Scenario:0x000000000c86ba80>
Did you mean?  lines
  /builds/gitlab-org/gitlab-ce/features/support/env.rb:52:in `before_scenario_run'

In Rails 5.0 `assert_template` and `assigns()` have been extracted to
the `rails-controller-testing` gem.
https://github.com/rails/rails/pull/20138/files

This commit adds this gem for rails5.

1. Synchronize used version of `parser` gem for both versions of rails: 4 and 5.
2. Fix broken CI pipelines for rails5 branches.

The 2.5.0.4 version is removed from rubygems, so it's skipped.
https://github.com/whitequark/parser/issues/478

- only older ruby and activerecord support dropped

These warnings would show when running rspec:

```
warning: parser/current is loading parser/ruby23, which recognizes
warning: 2.3.5-compliant syntax, but you are running 2.3.6.
warning: please see https://github.com/whitequark/parser#compatibility-with-ruby-mri.
```

Related PRs:

* https://github.com/whitequark/parser/pull/422
* https://github.com/mbj/unparser/pull/79

Related to gitaly#1060

Given 1.9 has issues we're aware of, for now we're locking to 1.8.
Closes gitlab-org/gitaly#1059 and will probably be removed if 1.10 is
tested an deemed ok.

Closes gitaly#1046

36847 - add changelog

26039 - add merge request id


26039 change due to lingust source file updates

Closes gitaly#985