Currently `change_column` cannot drop default if `null: false` is
specified at the same time. This change fixes the issue.

```ruby
  # cannot drop default
  change_column "tests", "contributor", :boolean, default: nil, null: false

  # we need the following workaround currently
  change_column "tests", "contributor", :boolean, null: false
  change_column "tests", "contributor", :boolean, default: nil
```

Closes #26582

`DISTINCT` clause is applied inside aggregate function by
`operation_over_aggregate_column` if needed. Unneeded outside aggregate
function.

```ruby
  # Before
  author.unique_categorized_posts.count
  # => SELECT DISTINCT COUNT(DISTINCT "posts"."id") FROM "posts" INNER JOIN "categorizations" ON "posts"."id" = "categorizations"."post_id" WHERE "categorizations"."author_id" = ?  [["author_id", 2]]

  # After
  author.unique_categorized_posts.count
  # => SELECT COUNT(DISTINCT "posts"."id") FROM "posts" INNER JOIN "categorizations" ON "posts"."id" = "categorizations"."post_id" WHERE "categorizations"."author_id" = ?  [["author_id", 2]]
```

Closes #27615

Since using a `ActiveSupport::Deprecation::DeprecatedConstantProxy`
would prevent people from inheriting this class and extending it
from the `ActiveSupport::HashWithIndifferentAccess` one would break
the ancestors chain, that's the best option we have here.

Remove styling from `true` and `false`.

Originally `quoted_id` was used in legacy quoting mechanism. Now we use
type casting mechanism for that. Let's deprecate `quoted_id`.

Fixes #26556.

The native timestamp type in MySQL is different from datetime type.
Internal representation of the timestamp type is UNIX time, This means
that timestamp columns are affected by time zone.

```
> SET time_zone = '+00:00';
Query OK, 0 rows affected (0.00 sec)

> INSERT INTO time_with_zone(ts,dt) VALUES (NOW(),NOW());
Query OK, 1 row affected (0.02 sec)

> SELECT * FROM time_with_zone;
+---------------------+---------------------+
| ts                  | dt                  |
+---------------------+---------------------+
| 2016-02-07 22:11:44 | 2016-02-07 22:11:44 |
+---------------------+---------------------+
1 row in set (0.00 sec)

> SET time_zone = '-08:00';
Query OK, 0 rows affected (0.00 sec)

> SELECT * FROM time_with_zone;
+---------------------+---------------------+
| ts                  | dt                  |
+---------------------+---------------------+
| 2016-02-07 14:11:44 | 2016-02-07 22:11:44 |
+---------------------+---------------------+
1 row in set (0.00 sec)
```

[Arthur Nogueira Neves & Matthew Draper]

[ci skip]

The column information for the testings table was being cached
so clear the cache in the test teardown.

This issue is only appear when you try to call `define_attribute_method`
and passing a symbol in Active Record. It does not appear in isolation
in Active Model itself.

Before this patch, when you run `User.define_attribute_method :foo`, you
will get:

    NoMethodError: undefined method `unpack' for :foo:Symbol
        from activerecord/lib/active_record/attribute_methods/read.rb:28:in `define_method_attribute'
        from activerecord/lib/active_record/attribute_methods/primary_key.rb:61:in `define_method_attribute'
        from activemodel/lib/active_model/attribute_methods.rb:292:in `block in define_attribute_method'
        from activemodel/lib/active_model/attribute_methods.rb:285:in `each'
        from activemodel/lib/active_model/attribute_methods.rb:285:in `define_attribute_method'

This patch contains both a fix in Active Model and a test in Active
Record for this error.

This ensures multiple threads inside a transactional test to see consistent
database state.

When a system test starts Puma spins up one thread and Capybara spins up
another thread. Because of this when tests are run the database cannot
see what was inserted into the database on teardown. This is because
there are two threads using two different connections.

This change uses the statement cache to lock the threads to using a
single connection ID instead of each not being able to see each other.
This code only runs in the fixture setup and teardown so it does not
affect real production databases.

When a transaction is opened we set `lock_thread` to `Thread.current` so
we can keep track of which connection the thread is using. When we
rollback the transaction we unlock the thread and then there will be no
left-over data in the database because the transaction will roll back
the correct connections.

[ Eileen M. Uchitelle, Matthew Draper ]

Pointed out by @matthewd that the HWIA subclass changes the
AS scoped class and top-level HWIA hierarchies out from under
existing classes.

This reverts commit 71da3909, reversing
changes made to 41c33bd4.

This constant was kept for the sake of backward compatibility; it
is still available under `ActiveSupport::HashWithIndifferentAccess`.

Furthermore, since Ruby 2.5 (https://bugs.ruby-lang.org/issues/11547)
won't support top level constant lookup, people would have to update
their code anyway.

Extra `data_source_exists?(table_name)` is unneeded if
`drop_table(table_name, if_exists: true)` directly.

Oracle database itself does not have `bigint` SQL type, then it gets `ORA-00902: invalid datatype`.

It can be addressed by using ActiveRecord `bigint` type
because Oracle enhanced adapter recognizes ActiveRecord `bigint` type
and transfer it to its equivalent SQL type `NUMBER(19)`.

Fixes CI failure caused by #27961.

https://travis-ci.org/rails/rails/jobs/201472146#L2390-L2409

primary_keys(table) needs to query various metadata tables in Postgres
to determine the primary key for the table. Previously, it did so using
a complex common table expression against pg_constraint and
pg_attribute.

This patch simplifies the query by using information_schema tables.
This simplifies the logic, making the query far easier to understand,
and additionally avoids an expensive unnest, window function query, and
common table expression.

Add `default_index_type?` to the generic schema dumper doesn't have the knowledge about an index type

The `default` arg of `index_name_exists?` is only used the adapter does
not implemented `indexes`. But currently all adapters implemented
`indexes` (See #26688). Therefore the `default` arg is never used.

Closes #27980

Closes #27979