`binds` is an array of a query attribute since Active Record 5.0.

Because `connection#to_sql` does not mutate `binds`.

Address to https://github.com/rails/rails/commit/5a302bf553af0e6fedfc63299fc5cd6e79599ef3#commitcomment-18288388.

Changing the order of method chaining `merge` and other query
method such as `joins` should produce the same result.

```ruby
class Topic < ApplicationRecord
  scope :safe_chaininig,   -> { joins(:comments).merge(Comment.newest) }
  scope :unsafe_chaininig, -> { merge(Comment.newest).joins(:comments) } #=> NoMethodError
end
```

Fixes #22398.

Albeit the previous existence of this method is not seen in
the patch itself, the declaration

    mattr_accessor :error_on_ignored_order_or_limit, instance_writer: false

was present before. It was removed recently in 210012f6.

- It adds lot of warnings while running the tests. After https://github.com/rails/rails/commit/451437c6f57e66cc7586ec966e530493927098c7 it is not needed.
- Added reader method with deprecation warning for backward compatibility.

Refactored method `ActiveRecord::ConnectionAdapters::PostgreSQL::SchemaStatements#default_sequence_name`

Follow up to #20818.

`retrieve_connection` is passed `spec_name` instead of `klass` since #24844.

Keeps schema migration inserts as a single commit, so we still get all of the
benefits of https://github.com/rails/rails/commit/42dd2336b31a8d98776d039a2b9fd7f834156a78,
but allows for easier git diff-ing. Fixes #25504.

We are setting a limit unconditionally just below,
which overrides any existing one anyway.

loaded model classes have their connections wrapped in transactions.

See #17776

In Rails 4 config.eager_load was changed to false in the test environment. This
means that model classes that connect to alternate databases with
establish_connection are not loaded at start up. If use_transactional_fixtures
is enabled, transactions are wrapped around the connections that have been
established only at the start of the test suite. So model classes loaded later
don't have transactions causing data created in the alternate database not to
be removed.

This change resolves that by creating a new connection.active_record
notification that gets fired whenever a connection is established. I then added
a subscriber after we set up transactions in the test environment to listen for
additional connections and wrap those in transactions as well.

When `group` is used in combination with any calculation method, the
resulting hash uses the grouping expression as the key. Currently we're
incorrectly always favoring the type reported by the query, instead of
the type known by the class. This causes differing behavior depending on
whether the adaptor actually gives proper types with the query or not.
After this change, the behavior will be the same on all adaptors -- we
see if we know the type from the class, fall back to the type from the
query, and finally fall back to the identity type.

Fixes #25595

The documentation states that parameter values longer than 20 characters
will be truncated by words, but the example shows that a parameter based
on "David Heinemeier Hansson" (with id: 125) becomes "125-david" when
"David Heinemeier".length == 16 so why so short?

The answer lies in the use of the #truncate option omission: nil which
seems to have been intended to mean "nothing", but which actually causes
the default string "..." to be used. This causes #truncate to cleave
words until the "..." can be added and still remain within the requested
size of 20 characters.

The better option is omission: '' (which is probably what was originally
intended).

Furthermore, since the use of #parameterize will remove non-alphanumeric
characters, we can maximize the useful content of the output by calling
parameterize first and then giving truncate a separator: /-/ rather than
a space.

This includes minor grammar fixes throughout the file.

This way, we aren't racing other threads, so we don't need to re-check
the conditional. And we no longer need to hold the lock while calling
remove (which can choose to make a new connection while we wait).

A concurrent thread may have also detected it to be stale, and already
released (or even reassigned) it by now.

Fixes #25585

A thread can only release a connection if it owns it, or it's owned by a
thread that has died.

This nested if checked the same value as the containing case statement.
Moved the code in the if/else into when/else in the containing case.

The current error message only indicates that a touch can fail because the record is new. In practice, we saw cases where touches were failing because the record had been destroyed. `persisted?` checks `new_record?` *and* `destroyed?`. It was confusing to get a message about a new record when in reality we were destroying records.

I also included a helpful tip for users to consider using `persisted?`, `new_record?`, or `destroyed?` before touching.

Actually schema dumper/creation supports composite primary key (#21614).
Therefore it should not show the warning about composite primary key in
connection adapter.

This change moves the warning to `AttributeMethods::PrimaryKey` and
suppress the warning for habtm join table.

Fixes #25388.

`create_table_info_cache` is used for sharing `create_table_info` both
`table_options` and `foreign_keys`. But `foreign_keys` no longer uses
`create_table_info_cache` by #25307. No need caching anymore.

This reverts commit 407e0ab5.

Because causing an extra query by `sql_for_insert` for guessing a
primary key.
https://github.com/rails/rails/blob/v5.0.0/activerecord/lib/active_record/connection_adapters/postgresql/database_statements.rb#L121-L125

Do not specal case inspecting associated arrays of over 10 elements, preventing infinite looping in some cases.