Make dependencies.rb add a name to NameError

The :singleton-method: directive is printed in the docs.
This directive is not necessary as the method is documentable just fine.

Add to guides app and helper objects on command line

Fix state leak.

Remove extra list item element in engine guide

Cleaning up the list items that detail what an engines `--full` option
generates. Joining a multiline multi list item into a multiline single
list item.

[ci skip]

Don't save through records twice

Conflicts:
	activerecord/CHANGELOG.md
	activerecord/test/cases/associations/has_many_through_associations_test.rb

ErrorsController shouldn't inherit ApplicationController, but ActionController

Always update counter caches in memory when adding records

Inheriting `ApplicationController` often causes an issue as it always
has before/after actions. It should encourage to use `ActionController::Base` instead.

[ci skip]

Don't use `Column` for type casting in Relation calculations

Don't use column object when calculating type cast defaults

Don't use column object for type casting in `quoting`

Don't rely on the column for type casting reflections

Further simplify `changed?` conditional for numeric types

remove unnecessary gsub for space in ActionView::Helpers#debug

Introduce an object to aid in creation and management of `@attributes`

Mostly delegation to start, but we can start moving a lot of behavior in
bulk to this object.

[ci skip]

Create custom errors page on ActionController guides

highlight/except support for regular expressions and blocks.

The helper will yield each matched word, and you can use this instead of the
':highlighter' option for more complex replacing logic:

  highlight('My email is me@work.com', EMAIL_REGEXP) { |m| mail_to(m) }
  # => 'My email is <a href="mailto:me@work.com">me@work.com</a>'

Don't include inheritance column in the through_scope_attributes

Now the following case will work fine

    class Tag < ActiveRecord::Base
    end

    class Publisher::Article < ActiveRecord::Base
      has_and_belongs_to_many :tags
    end

Fixes #15761

Fix typo in 4.1 upgrade/flash structure

Relpace `=~ Regexp.new str` with `.include? str` in `_valid_action_name?`

Remove unused block parameter from #gsub call in AM::MailHelper#block_format

Add `logger.debug?` guard to `ActionMailer::LogSubscriber#process`

*TLDR*: The method is 4 times faster when log level is higher than DEBUG.
Also, the other two methods, `#deliver` and `#receive` have similar guard statements,
so this commit adds some symmetry to the code.

This is probably not the most critical part of ActionMailer in terms
of performance, but here are some benchmarks:

```
require 'benchmark/ips'
require 'action_mailer'

event = ActiveSupport::Notifications::Event.new(
            'process.action_mailer',
            Time.now,
            Time.now,
            'bf4e2b36ce085fd35b24',
            { mailer: "UserMailer", action: :welcome }
        )

ActionMailer::Base.logger = ActiveSupport::Logger.new '/dev/null'

subscriber = ActionMailer::LogSubscriber.new
def subscriber.process_with_guard(event)
  return unless logger.debug?
  mailer = event.payload[:mailer]
  action = event.payload[:action]
  debug("\n#{mailer}##{action}: processed outbound mail in #{event.duration.round(1)}ms")
end

Benchmark.ips do |r|
  ActionMailer::Base.logger.level = ::Logger::Severity::INFO
  r.report('no guard') { subscriber.process(event) }
  r.report('   guard') { subscriber.process_with_guard(event) }
end

__END__
Calculating -------------------------------------
            no guard      9640 i/100ms
               guard     38381 i/100ms
-------------------------------------------------
            no guard   169166.9 (±10.2%) i/s -     838680 in   5.007262s
               guard   728184.9 (±9.6%) i/s -    3607814 in   4.999218s
```

When log level is DEBUG, the _guarded_ method is a bit slower. This is a
good tradeoff for 4x improvement in production mode.

```
Benchmark.ips do |r|
  ActionMailer::Base.logger.level = ::Logger::Severity::DEBUG
  r.report('no guard') { subscriber.process(event) }
  r.report('   guard') { subscriber.process_with_guard(event) }
end

__END__
Calculating -------------------------------------
            no guard      4970 i/100ms
               guard      4564 i/100ms
-------------------------------------------------
            no guard    55617.4 (±3.5%) i/s -     278320 in   5.010523s
               guard    49452.1 (±5.6%) i/s -     251020 in   5.093358s

```

`Type::Integer.new.type_cast('') # => nil`, we do not need a special
case to handle this, `nil => ''` already returns false. The only case we
need to handle is `0 => 'wibble'` should be changed, while `0 => '0'`
should not.

Because it is more natural way to test substring inclusion. Also, in
this particular case it is much faster.

In general, using `Regexp.new str` for such kind of things is dangerous.
The string must be escaped, unless you know what you're doing. Example:

    Regexp.new "\\" # HELLO WINDOWS
    # RegexpError: too short escape sequence: /\/

The right way to do this is escape the string

    Regexp.new Regexp.escape "\\"
    # => /\\/

Here is the benchmark showing how faster `include?` call is.

```
require 'benchmark/ips'

Benchmark.ips do |x|
  x.report('include?') { !"index".to_s.include? File::SEPARATOR }
  x.report('   !~   ') { "index" !~ Regexp.new(File::SEPARATOR) }
end

__END__
Calculating -------------------------------------
            include?     75754 i/100ms
               !~        21089 i/100ms
-------------------------------------------------
            include?  3172882.3 (±4.5%) i/s -   15832586 in   5.000659s
               !~      322918.8 (±8.6%) i/s -    1602764 in   4.999509s
```

Extra `.to_s` call is needed to handle the case when `action_name` is
`nil`. If it is omitted, some tests fail.

remove unused method from Postgresql::Adapter