Example:

    create_table :barcodes, primary_key: ["region", "code"] do |t|
      t.string :region
      t.integer :code
    end

Some test cases are testing only mysql adapter. We should test mysql2
adapter also.

`@connection` in `StatementPool` is only used for PG adapter.
No need for abstract `StatementPool` class.

The thread_safe gem is being deprecated and all its code has been merged
into the concurrent-ruby gem. The new class, Concurrent::Map, is exactly
the same as its predecessor except for fixes to two bugs discovered
during the merge.

Is this supposed to be public API? If so, I can document it instead.
📝

Other public APIs do not have these annotations.

[ci skip]

The schema_migrations table name is configurable. We should use this value when checking for ignored table names when dumping schema instead of a hardcoded value.

Follow up #21591.

The document of limit option for a text column is incorrect.

MySQL: the limit is byte length, not character length
Pg, Sqlite3: variable unlimited length

`restrict_with_error` message will now respect owner’s human name in any locale [kuboon & Ronak Jangir]

autoloading this could possibly cause some weird race condition
when calling an AR::Attribute's singleton method on a threaded server.

Bumps from `5.6` to `5.7`

The last call site of `last_version` was removed with:

838e1832

This method is private API and never used. Let's remove it.

`Schema#migrations_paths` is not supposed to be public API. In fact
it's only used inside `Schema` itself, so let's make it private.

It is always passed in

Usage was removed in 5c449553

This reverts commit d5ba9a42, reversing
changes made to 30c50339.

Reason: This generate the dynalic finders more than one time

[ci skip]

@sgrif can you review when you have time? Thanks!

These 3 methods expect `ConnectionAdapters` to have `tables` method,
so make it clear that `tables` method is interface.

* `ConnectionAdapters::SchemaCache#prepare_tables`
* `db:schema:cache:dump` task
* `SchemaDumper#tables`

Basically view tests for MySQL are same with
`test/cases/adapters/postgresql/view_test.rb`.

So move `test/cases/adapters/postgresql/view_test.rb` to
`test/cases/view_test.rb` and make them only run if
`current_adapter` supports writable view.

Its value never change since associations are defined at class load time
so there is no need to build the hash everytime the method is called.

Before this change:

    Calculating -------------------------------------
             reflections   804.000  i/100ms
    -------------------------------------------------
             reflections      8.213k (±26.2%) i/s -     36.180k

After this change:

    Calculating -------------------------------------
             reflections    24.548k i/100ms
    -------------------------------------------------
             reflections      1.591M (±25.7%) i/s -      7.364M

Benchmark script:

    require 'bundler/setup'

    require 'active_record'
    require 'benchmark/ips'

    ActiveRecord::Base.establish_connection(adapter: 'sqlite3', database: ':memory:')
    ActiveRecord::Migration.verbose = false

    ActiveRecord::Schema.define do
      100.times do |i|
        create_table "users#{i}", force: true
      end

      create_table :cars, force: true do |t|
        100.times do |i|
          t.references "users#{i}"
        end
      end
    end

    class Car < ActiveRecord::Base
      100.times do |i|
        belongs_to "users#{i}".to_sym
      end
    end

    Benchmark.ips do |x|
      x.report('reflections') { Car.reflections }
    end

Benchmark Script Used:
```
begin
  require 'bundler/inline'
rescue LoadError => e
  $stderr.puts 'Bundler version 1.10 or later is required. Please update your Bundler'
  raise e
end

gemfile(true) do
  source 'https://rubygems.org'
  gem 'rails', path: '~/rails' # master against ref "f1f0a3f8"
  gem 'arel', github: 'rails/arel', branch: 'master'
  gem 'rack', github: 'rack/rack', branch: 'master'
  gem 'sass'
  gem 'sprockets-rails', github: 'rails/sprockets-rails', branch: 'master'
  gem 'sprockets', github: 'rails/sprockets', branch: 'master'
  gem 'pg'
  gem 'benchmark-ips'
end

require 'active_record'
require 'benchmark/ips'

ActiveRecord::Base.establish_connection('postgres://postgres@localhost:5432/rubybench')

ActiveRecord::Migration.verbose = false

ActiveRecord::Schema.define do
  create_table :users, force: true do |t|
    t.string :name, :email
    t.timestamps null: false
  end
end

class User < ActiveRecord::Base; end

attributes = {
  name: "Lorem ipsum dolor sit amet, consectetur adipiscing elit.",
  email: "foobar@email.com",
}

1000.times { User.create!(attributes) }

Benchmark.ips(5, 3) do |x|
  x.report('where with hash single') { User.where(name: "Lorem ipsum dolor sit amet, consectetur adipiscing elit.") }
  x.report('where with string single') { User.where("users.name = ?", "Lorem ipsum dolor sit amet, consectetur adipiscing elit.") }
  x.report('where with hash double') { User.where(name: "Lorem ipsum dolor sit amet, consectetur adipiscing elit.", email: "foobar@email.com") }
  x.report('where with string double') { User.where("users.name = ? AND users.email = ?", "Lorem ipsum dolor sit amet, consectetur adipiscing elit.", "foobar@email.com") }
  x.compare!
end
```

Before:
```
Calculating -------------------------------------
where with hash single
                         3.300k i/100ms
where with string single
                         4.965k i/100ms
where with hash double
                         2.594k i/100ms
where with string double
                         4.400k i/100ms
-------------------------------------------------
where with hash single
                         35.161k (± 1.2%) i/s -    178.200k
where with string single
                         53.368k (± 2.9%) i/s -    268.110k
where with hash double
                         27.364k (± 1.1%) i/s -    137.482k
where with string double
                         46.876k (± 2.1%) i/s -    237.600k

Comparison:
where with string single:    53368.1 i/s
where with string double:    46875.5 i/s - 1.14x slower
where with hash single:    35160.8 i/s - 1.52x slower
where with hash double:    27364.0 i/s - 1.95x slower
```

After:
```
Calculating -------------------------------------
where with hash single
                         3.403k i/100ms
where with string single
                         5.167k i/100ms
where with hash double
                         2.659k i/100ms
where with string double
                         4.597k i/100ms
-------------------------------------------------
where with hash single
                         36.410k (± 1.3%) i/s -    183.762k
where with string single
                         55.009k (± 2.6%) i/s -    279.018k
where with hash double
                         27.951k (± 1.4%) i/s -    140.927k
where with string double
                         48.362k (± 2.6%) i/s -    243.641k

Comparison:
where with string single:    55008.6 i/s
where with string double:    48361.5 i/s - 1.14x slower
where with hash single:    36410.1 i/s - 1.51x slower
where with hash double:    27950.9 i/s - 1.97x slower
```

Known failure on Ruby 2.3/trunk: brianmario/mysql2#671

Example:

```ruby
class Person
  include ActiveModel::Validations

  attr_reader :name, :title
  validates_presence_of :name, on: :create
  validates_presence_of :title, on: :update
end

person = Person.new
person.valid?([:create, :update])    # => true
person.errors.messages               # => {:name=>["can't be blank"], :title=>["can't be blank"]}
```

Related with #20418