After 9d378747 `ActionDispatch::IntegrationTest`
class is loaded and defined in all Rails environments, not only test but also
production. This is not-intended loading of a class which is only used in
test environment.
To prevent not-intended loading, add `ActiveSupport.run_load_hooks` to
`ActionDispatch::IntegrationTest` with `action_dispatch_integration_test` name
and use it in `ActionMailer`.

This reverts changes made to integration tests in PR #23807.
The issue happens when using capybara with a driver that needs to start a
server in a separate thread like (poltergeist, selenium, etc).
Both threads the capybara server one and the test thread end running
syncronize over the interlock.

These should allow external code to run blocks of user code to do
"work", at a similar unit size to a web request, without needing to get
intimate with ActionDipatch.

Yesterday, when improving how `parsed_body` extracted a parser I wrote
77bbf1e9. Then I thought that was too many changes in one commit
and broke it up locally... or so I thought.

When pushed the extra commits removed the changes! Wups!

In shame, lob those changes together here:

  * 3b94c38a which meant to fix the CHANGELOG syntax error.
  * 5007df5e which meant to mention `parsed_body` in the docs.
  * 036a7a03 which meant to memoize the `parsed_body`.

Little easier to understand when you know the method that's used.

We're not guaranteed to have a `RequestEncoder` to assign on `get` requests
because we aren't extracting the parser from the response content type.

Until now.

When testing:

```ruby
post articles_path, params: { article: { title: 'Ahoy!' } }, as: :json
```

It's common to want to make assertions on the response body. Perhaps the
server responded with JSON, so you write `JSON.parse(response.body)`.
But that gets tedious real quick.

Instead add `parsed_body` which will automatically parse the reponse
body as what the last request was encoded `as`.

Prototype, you have served us well. But you are no longer how we make an
XMLHttpRequest. RIP

Turns

```
post articles_path(format: :json), params: { article: { name: 'Ahoy!' } }.to_json,
  headers: { 'Content-Type' => 'application/json' }
```

into

```
post articles_path, params: { article: { name: 'Ahoy!' } }, as: :json
```

We want to get rid of the `Live::Response` so we are consolidating methods
from `Live::Response` and `Response` by merging them together.

This adds an `#empty` method to the request so we don't need to
hard-code the empty array each time we call an empty
`ActionDispatch::Request`.

The work here is a continuation on combining controller and integration
test code bases into one.

`skip_filter`, `skip_action_callback` may both are deprecated in Rails 5.1 so waring msg should be specific.

Rails 4.x and earlier didn't support `Mime::Type[:FOO]`, so libraries
that support multiple Rails versions would've had to feature-detect
whether to use `Mime::Type[:FOO]` or `Mime::FOO`.

`Mime[:foo]` has been around for ages to look up registered MIME types
by symbol / extension, though, so libraries and plugins can safely
switch to that without breaking backward- or forward-compatibility.

Note: `Mime::ALL` isn't a real MIME type and isn't registered for lookup
by type or extension, so it's not available as `Mime[:all]`. We use it
internally as a wildcard for `respond_to` negotiation. If you use this
internal constant, continue to reference it with `Mime::ALL`.

Ref. efc6dd55

We should be asking the mime type method for the mime objects rather
than via const lookup

Actions are processed through `dispatch`, so they should have the
request set on them before any user land code can be executed.  Lets
stop setting _env on the controller, and give access to it through the
`env` method.

 - webservers do not do it
 - it makes redirect urls ugly when request.host is used for redirection

It's better to use Ruby methods when possible over methods defined by
Active Support because then it does not need to rely on any
dependencies.

Based on #19976 and #18305.

`*args` is not required here and should be avoided when not necessary
because `*args` are slower than `args` and create unnecessary array
allocations.

This method is being used in `#xml_http_request`, but was not properly
required. This causes `NoMethodError` on projects that are doing
integration test.

This fixes the reasons 4cf3b8ac, 303567e5, and fa634484 needed to be
reverted in 71420598. The revert has been reverted and this fixes
the issues caused previously.

If we call `super` first we will end up nuking the session settings in the
application tests that do `setup do` - so any session login or cookie
settings will not be persisted thoughout the test sessions.

Calling `super` last prevents `@integration_session` from getting nuked
and set to nil if it's already set.

Test added to prevent regression of this behavior in the future.

This reverts commit 71420598.

Haven't diagnosed yet. No similarly failing tests in Rails to work from.

cc @tenderlove, @eileencodes

Revert "there is always an integration session, so remove the check"
Revert "lazily create the integration session"
Revert "use before_setup to set up test instance variables"

This reverts commits 4cf3b8ac, 303567e5, and fa634484.

Fixes regression in #18423. Merge default headers for new responses,
but don't merge when creating a response from the last session request.

hat tip @senny ❤

there's really no benefit here.  It's the same number of lines without
the meta programming and is faster

now we don't have to call reset! everywhere

We shouldn't cache if it's not absolutely necessary. Removes
route caching and instead skips using the `url_helpers` is the
integration test session doesn't require it. Benchmark ips on
integration and controller index method tests below.

Without any caching or changes to `#url_helpers`:
```
Calculating -------------------------------------
INDEX: Integration Test
                        71.000  i/100ms
INDEX: Functional Test
                        99.000  i/100ms
-------------------------------------------------
INDEX: Integration Test
                        728.878  (± 8.0%) i/s -      3.692k
INDEX: Functional Test
                          1.015k (± 6.7%) i/s -      5.148k

Comparison:
INDEX: Functional Test:     1015.4 i/s
INDEX: Integration Test:      728.9 i/s - 1.39x slower
```

With caching on `#url_helpers`:
```
Calculating -------------------------------------
INDEX: Integration Test
                        74.000  i/100ms
INDEX: Functional Test
                        99.000  i/100ms
-------------------------------------------------
INDEX: Integration Test
                        752.377  (± 6.9%) i/s -      3.774k
INDEX: Functional Test
                          1.021k (± 6.7%) i/s -      5.148k

Comparison:
INDEX: Functional Test:     1021.1 i/s
INDEX: Integration Test:      752.4 i/s - 1.36x slower
```

Afer removing the caching and bypassing the `url_helpers` when not
necessary in the session:
```
Calculating -------------------------------------
INDEX: Integration Test
                        87.000  i/100ms
INDEX: Functional Test
                        97.000  i/100ms
-------------------------------------------------
INDEX: Integration Test
                        828.433  (± 6.4%) i/s -      4.176k
INDEX: Functional Test
                        926.763  (± 7.2%) i/s -      4.656k

Comparison:
INDEX: Functional Test:      926.8 i/s
INDEX: Integration Test:      828.4 i/s - 1.12x slower
```

in `ActionController::TestCase` and
`ActionDispatch::Integration`

Old syntax:

    `xhr :get, :create, params: { id: 1 }`

New syntax example:

    `get :create, params: { id: 1 }, xhr: true`

Non-kwargs requests are deprecated now.
Guides are updated as well.

`post url, nil, nil, { a: 'b' }` doesn't make sense.
`post url, params: { y: x }, session: { a: 'b' }` would be an explicit way to do the same