Fix CI error with minitest 5.14

rack 2.1.0 changed expires date format from rfc2822 to httpdate

Remove Ruby 2.7 warning on arg forwarding

https://github.com/seattlerb/minitest/commit/5802aa08280a3609b684e0ebef95dabfbd3ec9ff
https://github.com/seattlerb/minitest/pull/809#discussion_r332399750

To support both 5.13.0 and 5.14.0, use `#initialize` instead.

Add AC::TestSession#dig method like AD::Request::Session

In 2aa33bbd I left out a change from the
original PR (#38204) when I did my backport. This will fix the test
failures on CI.

* Fix ActionController::TestSession#id to return Rack::Session::SessionId instance

* test SessionId#public_id

* test session["session_id"]
Co-authored-by: NBenjamin Quorning <22333+bquorning@users.noreply.github.com>

Manual backport of 2c07d162.

Since test fixtures share connections (due to transactional tests) we
end up overwriting the reading configuration so Rails doesn't recognize
it as a replica connection.

This change ensures that if we're using the `reading` role that
connections will always have prevent writes turned on.

If you need a replica connection that does not block writes, you should
use a different role name other than `:reading`.

The db selector test and connection handlers test have been updated to
test for these changes. In the db selector test we don't always have a
writing handler so I updated test fixtures to return if that's nil.

Lastly one test needed to be updated to use a different handler name due
to it needing to write to successfully test what it needs to test.

Fixes #37765

This reverts commit 6568ca66, reversing
changes made to 7cad464e.

Closes #38187.

The main interface to eager loading is config.eager_load. The logic that
implies happens during the boot process.

With the introduction of Zeitwerk, application code is loaded in the
finisher as everything else, but in previous versions of Rails users
could eager load the application code regardless of config.eager_load.

Use cases:

   * Some gems like indexers need to have everything in memory and would
   be a bad user experience to ask users to conditionally set the eager
   load flag.

   * Some tests may need to have everything in memory and would be a bad
   experience to have the flag enabled globally in the test environment.

I personally feel that the contract between this method and the entire
eager loading process is ill-defined. I believe this method is
essentially internal. The purpose of this patch is simply to restore this
functionality emulating what it did before because rethinking the design
of this interface may need time.

Fix collision check when using a generator and using the force option

Only assign @new_record_before_save once in autosave_association

Fix random CI fail due to non-deterministic order

https://buildkite.com/rails/rails/builds/66168#b348f446-0406-425e-ba53-c5d15034cf98/1002-1070

Fix adding non-null column to existing SQLite3 table

Avoid assigning duplicate values to join_values

Use travel_to to in date_helper_test.rb.

Update "The Hard Way" setup for Arch Linux

Fix regression, enforce fresh ETag header after collection contents change

Add `ActionDispatch::HostAuthorization` to "Configuring Middleware" guide [ci skip]

Separate with space between table name and columns list in SQL of insert_all

Return unfrozen strings from to_sentence

Do not abort `rails new` when git binary doesn't exist

I misunderstood the effect of `ruby2_keywords`, the effect is not
local scope, the effect still alive even where you delegate.