Deprecate aggregations with group by duplicated fields

We've learned that `merge` causes duplicated multiple values easily, so
if we missed to deduplicate the values, it will cause weird behavior
like #38052, #39171.

I've investigated the deduplication for the values, at least that had
existed since Rails 3.0.

https://github.com/rails/rails/commit/bed9179aa1496f6d28891cf515af0d7e515ebbab

Aggregations with group by multiple fields was introduced at Rails 3.1,
but we missed the deduplication for the aggregation result, unlike the
generated SQL.

https://github.com/rails/rails/commit/a5cdf0b9eb860c4370ae5fde231e1b61f71b6b65

While the investigation, I've found that `annotate` is also missed the
deduplication.

I don't suppose this weird behavior is intended for both.

So I'd like to deprecate the duplicated behavior in Rails 6.1, and will
be deduplicated all multiple values in Rails 6.2.

To migrate to Rails 6.2's behavior, use `uniq!(:group)` to deduplicate
group fields.

```ruby
accounts = Account.group(:firm_id)

# duplicated group fields, deprecated.
accounts.merge(accounts.where.not(credit_limit: nil)).sum(:credit_limit)
# => {
#   [1, 1] => 50,
#   [2, 2] => 60
# }

# use `uniq!(:group)` to deduplicate group fields.
accounts.merge(accounts.where.not(credit_limit: nil)).uniq!(:group).sum(:credit_limit)
# => {
#   1 => 50,
#   2 => 60
# }
```

- Language changes around the deliver_later note.
- Leave out "now", this is not a new feature anymore
- Drops lots repetitive "just" and "simple"

Deduplicate same clauses in `merge`

Fix through association with source/through scope which has joins

Related to #39328, #39358.

For now, `merge` cannot override non-equality clauses, so non-equality
clauses will easily be duplicated by `merge`.

This deduplicates redundant same clauses in `merge`.

Test Attached::Many in Attached::Many test

This exact test exists in `one_test.rb`.  It was probably overlooked
after being copied and pasted.

Reduce allocations in action_view cache expiry

Allow Erubi bufvar to be configured

This reverts 8538dfdc, which broke the
activemodel-serializers-xml gem.

We can still get most of the benefit by applying the optimisation from
7b391977 to empty hashes as well as nil.
This has the additional benefit of retaining the optimisation when the
user passes an empty options hash.

This allows `render_in` components that compile their own templates to
use their view context's current output buffer while a
`with_output_buffer` block is being evaluated.

Partially addresses #39377.

If source/through scope references other tables in where/order, we
should explicitly maintain joins in the scope, otherwise association
queries will fail with referenced unknown column.

Fixes #33525.

They may be cached for existing apps upgrading to 6.1. Follows up on #34477.

Fix through association to respect source scope for `includes`/`preload`

Decoding JSON dates should respect newlines

This fixes tests for missing backtraces

Errors *should* output their stack. Somehow these tests were passing
before. This clarifies all states (skip/fail/error) and works against
@tenderlove's patch.

`reflection.scope` is not aware of all source scopes if the association
is through association.

It should use `reflection.join_scopes` for that.

Fixes #39376.

Updates to nokogiri 1.10.9

Add `ActionDispatch::ActionableExceptions` to default middleware docs

Before #39304 was merged, backtraces would look like this:

```
$ be ruby -I lib:test test/controller/request_forgery_protection_test.rb -n test_ignores_trailing_slash_during_generation
Run options: -n test_ignores_trailing_slash_during_generation --seed 22205

E

Error:
PerFormTokensControllerTest#test_ignores_trailing_slash_during_generation:
NoMethodError: undefined method `have_cookie_jar?' for #<Object:0x00007f9ce19b9620>
    /Users/aaron/git/rails/actionpack/lib/action_controller/test_case.rb:522:in `ensure in process'
    /Users/aaron/git/rails/actionpack/lib/action_controller/test_case.rb:542:in `process'
    /Users/aaron/git/rails/actionpack/lib/action_controller/test_case.rb:395:in `get'
    test/controller/request_forgery_protection_test.rb:1005:in `test_ignores_trailing_slash_during_generation'

rails test test/controller/request_forgery_protection_test.rb:1004

Finished in 0.213230s, 4.6898 runs/s, 0.0000 assertions/s.
1 runs, 0 assertions, 0 failures, 1 errors, 0 skips
```

After #39304 was merged they look like this:

```
$ be ruby -I lib:test test/controller/request_forgery_protection_test.rb -n test_ignores_trailing_slash_during_generation
Run options: -n test_ignores_trailing_slash_during_generation --seed 62892

E

Error:
PerFormTokensControllerTest#test_ignores_trailing_slash_during_generation:
NoMethodError: undefined method `have_cookie_jar?' for #<Object:0x00007fa6d60193c8>

rails test test/controller/request_forgery_protection_test.rb:1004

Finished in 0.211953s, 4.7180 runs/s, 0.0000 assertions/s.
1 runs, 0 assertions, 0 failures, 1 errors, 0 skips
```

This patch reverts the part of #39304 that removes the backtrace.

Fixes #39088.

Resolve conflict between counter cache and optimistic locking

Satisfy Rubocop

Fixes offense:

```
actionpack/test/journey/router_test.rb:493:37: C: Style/HashSyntax: Use the new Ruby 1.9 hash syntax.
            options = options.merge(:_recall => recall)
                                    ^^^^^^^^^^^
```

Bump an Active Record instance's lock version after updating its counter
cache. This avoids raising an unnecessary ActiveRecord::StaleObjectError
upon subsequent transactions by maintaining parity with the
corresponding database record's lock_version column.

Journey is considered internal, and I would like to stop testing
internals so that we are free to refactor.  This commit just moves us up
one more call frame with the goal if getting closer to a public API.
Once we're up the stack high enough that we're testing a public API we
can stop

Unify raise_on_missing_translations for views and controllers

Add the configuration option for annotating templates with file names to the generated app

Clarify versioning strategy in maintenance policy

Not all releases are using `X.Y.Z`.

Since 7b391977, `serializable_hash`
allocates fewer objects when options is nil rather than empty.