Replace (slower) block.call with (faster) yield

This reverts commit 0ab075e7.

Performance optimization: `yield` with an implicit `block` is faster than `block.call`.
See http://youtu.be/fGFM_UrSp70?t=10m35s and the following benchmark:

```ruby
require 'benchmark/ips'

def fast
 yield
end

def slow(&block)
 block.call
end

Benchmark.ips do |x|
 x.report('fast') { fast{} }
 x.report('slow') { slow{} }
end

# => fast    154095 i/100ms
# => slow     71454 i/100ms
# =>
# => fast  7511067.8 (±5.0%) i/s -   37445085 in   4.999660s
# => slow  1227576.9 (±6.8%) i/s -    6145044 in   5.028356s
```

Performance optimization: `yield` with an implicit `block` is faster than `block.call`.
See http://youtu.be/fGFM_UrSp70?t=10m35s and the following benchmark:

```ruby
require 'benchmark/ips'

def fast
 yield
end

def slow(&block)
 block.call
end

Benchmark.ips do |x|
 x.report('fast') { fast{} }
 x.report('slow') { slow{} }
end

# => fast    154095 i/100ms
# => slow     71454 i/100ms
# =>
# => fast  7511067.8 (±5.0%) i/s -   37445085 in   4.999660s
# => slow  1227576.9 (±6.8%) i/s -    6145044 in   5.028356s
```

`cleanup_destination_root` method is not found anywhere.
Instead, `prepare_destination` clean up distination root on setup.

Closes #17197.
Closes #17207.

`{required}` is a type modifier so it should be:

    user:references{required}:index

and not:

    user:references:index{required}

It was changed to not set the cookie in get requests at
https://github.com/rails/turbolinks/commit/62cc3db457ad9745ccda047dd43ab84fb3d89707.

Related with dcb05f26.

It is temporary until we do the final release

Since we want this flag to be enabled anytime we are running the tests
under JRuby, let's enable this at the Rakefile level so people get the
performance boost on their local checkout.

Moreover, we avoid having to update this particular line anytime the
option changes on the JRuby side.

The only drawback is that we have to define it in every Rakefile but
there's no big deal, this is already the case for other options.

Right now BenchmarkCleaner allocates hundreds of strings on every request with an exception. This patch moves those strings to be generated at boot once and re-used.

## Bench Methods

I took a modified form of CodeTriage https://github.com/schneems/codetriage-ko1-test-app/blob/master/perf.rake and ran given rake tasks with and without the patch to BacktraceCleaner.

I made an endpoint results in exception

```
  def index
    raise “foo"
  end
```

The gem `memory_profiler` was used to capture objects allocated for a single request. Then `benchmark/ips` was used to test the speed of the patch.

You will see that we use less objects and the code becomes measurably faster with this patch.

## With patch:

Memory for one request:

```
Total allocated 7441
Total retained 37
```

Requests per second:


```
Calculating -------------------------------------
                 ips         4 i/100ms
-------------------------------------------------
                 ips       43.0 (±4.7%) i/s -        216 in   5.037733s
```


## Without patch:

Memory used for one request:


```
Total allocated 11599
Total retained 35 
```

Requests per second:

```
Calculating -------------------------------------
                 ips         3 i/100ms
-------------------------------------------------
                 ips       39.4 (±7.6%) i/s -        198 in   5.052783s
```

## Analysis

The patch is faster:

```
(43.0 - 39.4 ) / 39.4 * 100
# => 9 # % ~ speed bump
```

It also allocates less objects:

```
11599 - 7441
# => 4158
```

These strings are allocated on __EVERY SINGLE REQUEST__. This patch saves us 4158 objects __PER REQUEST__ with exception.

Faster errors == Faster applications

Also add to upgrading guide a notice about the deprecated sanitizers

fixture_path is a class attribute of ActiveSupport::TestCase, so
ActiveSupport::TestCase.method_defined?(:fixture_path=) would always
return false.

This should fix #4971

 * The assertions in AS::Duration don't actually assert.
 * The assertion in Railtie will pass even when `eager_load_namespaces`
   doesn't include `AppTemplate::Application` if `Rails.application` is
   truthy.

For more details, see here:

 * https://github.com/rails/rails/pull/16998
 * https://github.com/rails/rails/pull/17000

Fixes #16578

This seems to be the style settled on in most of the templates.

If code is not eager loaded constants are loaded on demand. Constant
autoloading is not thread-safe, so if eager loading is not enabled
multi-threading should not be allowed.

This showed up in certain Capybara scenarios: Most Capybara drivers
other than Rack::Test need a web server. In particular, drivers for
JavaScript support. Capybara launches WEBrick in its own thread for
those but that per se is fine, because the spec thread and the server
thread are coordinated.

Problem comes if the page being served in the spec makes Ajax calls.
Those may hit WEBrick in parallel, and since WEBrick is multi-threaded
and allow_concurrency? returns true in the test environment before
this patch, threads are spawned to serve those parallel requests. On
the other hand, since eager_load is false by default in the test
environment, constants are not preloaded.

So the suite is autoloading constants in a multi-threaded set. That's
a receipt for paracetamol. The symptom is random obscure errors whose
messages point somehow to constant autoloading.

As a consequence of this fix for allow_concurrency? WEBrick in
Capybara scenarios no longer runs in multi-threaded mode.

Fixes #15089.

* Fix long lines;
* Use simple quotes for string literals when interpolation is not used.