feat: Add single_line_empty_body config only_for_constructors

[rudiedirkx]

Don't fold anything, except a class' __construct(), because that looks better with PHP 8's property promotion.

I have plenty of empty methods that are overridden in children, or that are overrides of parents. I don't want to fold those, because that looks weird:

protected function applyConfig(int $priority, array $config, TreeAwareConfigContainer $parent): void {}

(where are the brackets?? am I in an interface?? help!!)

But empty PHP 8 constructors look weird if they're NOT folded:

public function __construct(
	protected string $name,
)
{

}

or

public function __construct(
	protected string $name,
) {
}

etc.

---

If the general idea is acceptable, it could be configured differently, with a context config option or something. Default is everything:

[
	'contexts' => ['class', 'interface', 'trait', 'enum', 'method', 'constructor', 'function'], // more?
],

[coveralls]

coverage: 96.055% (+0.002%) from 96.053%
when pulling 2763af9 on rudiedirkx:single_line_empty_body-only_for_constructors
into 5c6b9a4 on PHP-CS-Fixer:master.

[keradus]

Hi and thanks for your request.

Is this part of a standard or used by a bigger (F)OSS community? For us that is a condition to add a rule to this project. Otherwise, you can create a custom fixer and use that in your projects. Thanks for understanding.

[Wirone]

Personally I don't mind if this is part of a bigger community's standard. I just don't like the approach. IMHO it should be an option for configuring places where {} should be applied (constructor, method signature, loop, try/catch etc), by default all of them should be enabled, with ability to narrow the scope. This particular option is not flexible enough.

[keradus]

I'm challenging the need for proposed option as well as for generic option.
I do not want to have option for sth because "it is technically possible", but because it's a recommended standard.

Until proven differently (with link to example community that will benefit from given configuration), I'm against config for this rule.

Having one approach for given element (eg constructor), but another approach for another element - it's not a standard at all.

[rudiedirkx]

It's not a standard as far as I know. It's just how I like it.

[Wirone]

@keradus it was discussed in #7809, it really looks like making it configurable is something nice to have.