Don't use remove_method or remove_possible_method just before a new
definition: at best the purpose is unclear, and at worst it creates a
race condition.

Instead, prefer redefine_method when practical, and
silence_redefinition_of_method otherwise.

Besides making the code easier to read, this commit
also makes it faster:

* We don't eval `@base.class.respond_to?(:i18n_scope)` twice
* We only eval `@base.class.i18n_scope` once
* We don't call `flatten!` because it's not needed anymore
* We don't call `compact` because all elements are Symbols

[ci skip]

Closes #30012

Between 4.2 and 5.0 the behavior of how multiparameter attributes
interact with `_before_type_cast` changed. In 4.2 it returns the
post-type-cast value. After 5.0, it returns the hash that gets sent to
the type. This behavior is correct, but will cause an issue if you then
tried to render that value in an input like `text_field` or
`hidden_field`.

In this case, we want those fields to use the post-type-cast form,
instead of the `_before_type_cast` (the main reason it uses
`_before_type_cast` at all is to avoid losing data when casting a
non-numeric string to integer).

I've opted to modify `came_from_user?` rather than introduce a new
method for this as I want to avoid complicating that contract further,
and technically the multiparameter hash didn't come from assignment, it
was constructed internally by AR.

Close #27888.

Because `generated_attribute_methods` is an internal API.

ActiveModel::Errors#merge! allows ActiveModel::Errors to append errors from
a separate ActiveModel::Errors instance onto their own.

Example:

    person = Person.new
    person.errors.add(:name, :blank)

    errors = ActiveModel::Errors.new(Person.new)
    errors.add(:name, :invalid)

    person.errors.merge!(errors)
    puts person.errors.messages
    # => { name: ["can't be blank", "is invalid"] }

`ActiveModel::Type::DateTime#serialize` should return a `Time` object
so that finding by a datetime column works correctly.

This reverts commit 3420a145, reversing
changes made to afb66a5a.

[ci skip]

Fixes #29512.

Includes two external changes because they're referenced within the ActiveModel test suite.

The output of two string attributes is displayed differently in the docs. Standardize the output by always showing it as a comment.

* Allow a default value to be declared for class_attribute

* Convert to using class_attribute default rather than explicit setter

* Removed instance_accessor option by mistake

* False is a valid default value

* Documentation

a user input for a decimal column were ignored by numerically validations

[ci skip]

".. with __dir__ we can restore order in the Universe." - by @fxn

Related to 5b8738c2

When define callbacks latest definition on the same callback/method
overwrites previous ones.

* Remove trailing spaces.
* Add backticks around method and command.
* Fix indentation.

\Z was a mistake of \z. Replace \Z to \z to prevent newly \Z added.

Follow up of #17148.

We already have a _read_attribute method that can get the value we need
from the model.  Lets define that method in AM::Dirty and use the
existing one from AR::Dirty rather than introducing a new method.

[ci skip]

See 34321e4a for background on
ImmutableString vs String.

Our String type cannot delegate typecasting to ImmutableString, because
the latter freezes its input: duplicating the value after that gives us
an unfrozen result, but still mutates the originally passed object.

```ruby
nil.respond_to?(:stringify_keys) # => false
```

```ruby
include ActiveModel::AttributeAssignment
Object.new.assign_attributes(nil) #  => ArgumentError: When assigning attributes, you must pass a hash as an argument.
```

`messages` has `default_proc` so calling `reject` causes the warning.

https://github.com/ruby/ruby/blob/v2_4_1/hash.c#L1335-L1337