Drop jQuery as a dependency

Introduce `reload_<association>` reader for singular associations.

This patch brings back the functionality of passing true to the
association proxy. The behavior was deprecated with #20888 and scheduled
for removal in Rails 5.1.

The deprecation mentioned that instead of `Article.category(true)` one
should use `article#reload.category`. Unfortunately the alternative does
not expose the same behavior as passing true to the reader
did. Specifically reloading the parent record throws unsaved changes and
other caches away. Passing true only affected the association.

This is problematic and there is no easy workaround. I propose to bring
back the old functionality by introducing this new reader method for
singular associations.

Fix a wrong permission

2.3.2 has a known regression https://bugs.ruby-lang.org/issues/12920#note-5

form_helper typo fix [ci skip]

deep symbolize keys on secrets.yml

Implemented in 67f81cc7.

* Add form_with to unify form_tag/form_for.

`form_tag` and `form_for` serve very similar use cases. This
PR unifies that usage such that `form_with` can output just
the opening form tag akin to `form_tag` and can just work with
a url, for instance.

`form_with` by default doesn't attach class or id to the form —
removing them on fields is moved out to a default revisiting PR later.

Ported over old tests where applicable to ensure maximum coverage,
but left some commented out because they don't yet apply (e.g.
`fields_for` later being replaced by `fields`).

[ Kasper Timm Hansen & Marek Kirejczyk ]

* Add fields DSL method.

Strips `_for` and requires models passed as a keyword argument.

* Document form_with.

Graft the `form_for` docs: rewrite, revise and expand where
needed.

Also test that a `format` isn't used when an explicit URL
is passed.

* Enable remote by default.

Brand new world! Forms submit via XHRs by default, woah.

* Invert `include_id` to `skip_id`.

`skip_id: true` reads better than `include_id: false` (since the
`include_id` default is true).

* Invert `remote` to `local`.

Since forms are remote by default, the option name makes more sense
as `local: true`.

* Invert `enforce_utf8` to `skip_enforcing_utf8`.

* Refer to the brand spanking new rails-ujs.

Soon to be bundled in Rails proper, so jquery-ujs is out.

* Make `form_with` a bit more composed.

The flow is still not quite what it should be because the legacy
methods and these new ones pull at opposite ends.

Lots of options have been renamed, so now the new pieces don't fit
in so well.

I'll try to work on this in later commits after this PR (it's likely
there's a much better way to structure this whole part of Action View).

Update RELEASING_RAILS.md

I have removed the dot because only one string had a dot, all the other strings in the list had no dots

The test runner was updated to make use of autorun. This caused the
`bin/test` scripts to run Minitest twice.

Action Cable: JavaScript test updates

https://travis-ci.org/rails/rails/jobs/165766716#L506

https://github.com/rails/rails/commit/fc4c15d823b112cc440cbd9a4288e1a389c138da

When `ActiveRecord::Associations.eager_load!` is called, it eager loads
the `Preloader` but does not trigger a `eager_load!` on the `Preloader`
itself. A eager loaded component is responsible to forward the
`eager_load!` call to nested constants that make use of `eager_autoload`
as well.

This will also resolve the occasional `NameError` that we've observed in
production:

```
NameError: uninitialized constant ActiveRecord::Associations::Preloader::BelongsTo
```

Refactor ActiveSupport::XMLMini test suite

This fixes an error where the test runner would try and run
XMLMiniEngineTest like a normal test class, except it's abstract. Now,
to circumvent this, we don't include any of the actual tests in
XMLMiniEngineTest; they are wrapped in a module that is included in
subclass when they inherit from XMLMiniEngineTest. Pretty neat, huh?

Created a new module (a la Action Cable subscription adapter's test
suite) to be included in all sub class test to ensure compatability and
reduce duplicated code.

When making the new command insfrastructure I had missed that
`bin/rails runner some_file.rb some args` would pass the extra
args onto the file in `ARGV`.

Now fixed by allowing the command to take extra args again, and
make sure to remove the file name from `ARGV`.

Fix prepared statements disabled test again

- Due to `assert_nothing_raised` this test was not really testing
  anything.
- So updated it to assert that the query gives expected result.
- Also in general we can use `connection.unprepared_statement` for
  testing queries w/o prepared statements but it can't be used in this
  case. This test cases was added because when prepared_statements
  config is set to false, then DetermineIfPreparableVisitor module
  does not extended by Arel visitor resulting into an error. Ref: https://github.com/rails/rails/pull/22748.
- Because DetermineIfPreparableVisitor module does not get added to the
  visitor chain only if prepared_statements is false while **setting up
  connection**, not when `unprepared_statement` is used.
- I have also added an assertion for making sure that prepared_config
  is set to false from the start, so that nobody accidentally removes
  the connection setup using `arunit_without_prepared_statements` and
  replaces it with stubs or unprepared_statement.

Do not rely on Ruby implementation in AM test

Use existing --skip-turbolinks option for conditionals instead of checking if turbolinks is present in gemfile list

Now a few tests in ActiveModel rely on Ruby implementation and the fact
that in MRI `97.18` as a float is greater than `97.18` as a BigDecimal.

This is only relevant for MRI. On JRuby, comparing float to BigDecimal
would be conversion of them to the same type and they will be equal.

I'd like the ActiveModel test suite to be Ruby implementation-agnostic.
Here we test ActiveModel, not the Ruby internals.
This PR fixes a couple more JRuby tests.

Change html-scanner to rails-html-sanitizer

[ci skip]

Revert "fix typo in `match` doc [ci skip]"