This feature adds a lot of complication to ActiveRecord for dubious
value. Let's talk about what it does currently:

class Customer < ActiveRecord::Base
  composed_of :balance, :class_name => "Money", :mapping => %w(balance amount)
end

Instead, you can do something like this:

    def balance
      @balance ||= Money.new(value, currency)
    end

    def balance=(balance)
      self[:value] = balance.value
      self[:currency] = balance.currency
      @balance = balance
    end

Since that's fairly easy code to write, and doesn't need anything
extra from the framework, if you use composed_of today, you'll
have to add accessors/mutators like that.

Closes #1436
Closes #2084
Closes #3807

Based on examples seen in the Rails test suite. [ci skip]

Now isn't that better?

And yes, I know that private has no impact on class methods - it's a visual distinction, not a technical one.

Without this, some tasks won't load configurations
like custom engines.

Minimal implementation that supports db:create SQLite replacement

Get rid of ActiveModel::Configuration, make better use of
ActiveSupport::Concern + class_attribute, etc.

Before 7f4b0a1231bf3c65db2ad4066da78c3da5ffb01, this test are asserting
that update_attribute does the dirty tracking. Since we remove this
method and update_column write in the database directly this tests will
always fail>

Historically, update_attribute and update_attributes are similar, but
with one big difference: update_attribute does not run validations.
These two methods are really easy to confuse given their similar
names. Therefore, update_attribute is being removed in favor of
update_column.

See the thread on rails-core here:
https://groups.google.com/forum/?fromgroups#!topic/rubyonrails-core/BWPUTK7WvYA

That change to update_attribute was considered
to be too subtle and was reverted in 30ea9230
just before Rails 3 shipped. Later we introduced
update_column (Rails 3.1).

reflection.

ActiveRecord::FinderMethods#construct_limited_ids_condition will raise
ThrowResult if the limited reflection comes back empty. The other callers
of #construct_limited_ids_condition handle this exception (more specifically,
the callers of construct_relation_for*), but #exists? didn't until now.

This behavior was added in be4ecdcc.

Closes #1139.

Fixes #2553, #1141, #1623 and #2062.