nil or false should not be valid argument to the merge method.

Closes #12264

Relation#merge checks if the argument is an array and the only possible
returns of build_default_scope is nil or a Relation.

Doing this we can raise an ArgumentError when Relation#merge receive a
nil value.

Add an `:only` option to `perform_enqueued_jobs` to filter jobs based on type.

type.

This allows specific jobs to be tested, while preventing others from
being performed unnecessarily.

Example:

    def test_hello_job
      assert_performed_jobs 1, only: HelloJob do
        HelloJob.perform_later('jeremy')
        LoggingJob.perform_later
      end
    end

An array may also be specified, to support testing multiple jobs.

Example:

    def test_hello_and_logging_jobs
      assert_nothing_raised do
        assert_performed_jobs 2, only: [HelloJob, LoggingJob] do
          HelloJob.perform_later('jeremy')
          LoggingJob.perform_later('stewie')
          RescueJob.perform_later('david')
        end
      end
    end

Fixes #18802.

Trim space and document :only option.

The `&block` form is more than twice as fast as the manual form of
delegation (and is the code I'd rather write anyway). Unfortunately,
it's still twice as slow on MRI. However, this is enough of a hotspot to
justify giving JRuby special treatment.

I can't currently provide benchmarks in the context of Active Record,
since the JDBC adapters still aren't updated for 4.2, but the actual
work performed (assuming it's been read at least once already) will have
nearly the same performance characteristics as
https://gist.github.com/sgrif/b86832786551aaee74de.

We already validate the keys, so it is better to use the built-in
feature to do this

The same is not true of `define_attribute`, which is meant to be the low
level no-magic API that sits underneath. The differences between the two
APIs are:

- `attribute`
  - Lazy (the attribute will be defined after the schema has loaded)
  - Allows either a type object or a symbol
- `define_attribute`
  - Runs immediately (might get trampled by schema loading)
  - Requires a type object

This was the last blocker in terms of public interface requirements
originally discussed for this feature back in May. All the
implementation blockers have been cleared, so this feature is probably
ready for release (pending one more look-over by me).

Merge #12067 rebasing and improving the code.

Enable number_to_percentage to keep the number's precision by allowing :precision option value to be nil

Conflicts:
	activesupport/CHANGELOG.md
	activesupport/lib/active_support/number_helper.rb
	activesupport/test/number_helper_test.rb

Indicate link_to creates an anchor element

The `link_to` helper generates an HTML anchor element (consisting of opening and closing anchor tags and an element body). The docs currently state the a link tag is generated (which would indicate a tag like `<link>`, which is another valid HTML tag), so this change clarifies that an anchor element is actually generated.

[ci skip]

This reverts commit 9c47b874.

These are (potentially, depending on input) called in several places in
both the router, and Active Record. The code also becomes much cleaner.
This results in ~33% performance gain in both methods.

    Calculating -------------------------------------
                  before    15.696k i/100ms
                  after     19.865k i/100ms
    -------------------------------------------------
                  before    303.064k (± 2.6%) i/s -      1.523M
                  after     446.734k (± 2.4%) i/s -      2.245M

On Ruby 2.2, a warning will be emitted about states not being copied,
because we're calling `super` from a subclass. We can safely ignore it,
however, since we're converting the result back into a HWIDA

Amended json_escape comments

Better docs for AM::Lint::Tests

The existing example seems somewhat forced: is it realistic
to have a model that accepts state in its initializer but
considers it has not been changed? By allowing state changes
to happen only via accessors it seems more natural that new
instances are considered to be unchanged (as they are in AR).

[ci skip]

Use SCHEMA instead of DB_STRUCTURE for specifiying structure file.

Conflicts:
	activerecord/CHANGELOG.md

Fix `test_types_line_up` when column type missing

Extracted silence_stream method to new module in activesupport/testing

Migrating xhr methods to keyword arguments syntax

Also remove the default value since it will be always passed and
Array(nil) returns an empty array

Rename AS::Callback#merge

This will make rake test_sqlite3_mem work again

They are implementation details

Deprecated finders are not supported anymore

Closes #18646

call is too generic

Thse methods are used only inside this class

Thor isn't very discerning over whether some content is present when passed
to `inject_into_file`, e.g. a commented out route is detected as being present.
So to prevent people scratching their heads as to why a route hasn't appeared
it's better to fall on the side of having duplicate routes.

The `model_name` method should be called on `to_model`

because the partial renderer would not create an lvar per each template since c67005f2

config_accessor should better not be a public method, as with Ruby's attr_accessor