[TypeDeclaration] Skip return interface on NarrowObjectReturnTypeRector

[samsonasik]

Per @calebdw feedback on #7575 (comment)

This PR skip return interface, except for common use that usually use dummy/implementation on tests, defined in constant ALLOWED_INTERFACES_CHANGE list.

/cc @Orest-Divintari

[samsonasik]

New PHPStan release

https://github.com/phpstan/phpstan/releases/tag/2.1.32

seems cause unit test error:

https://github.com/rectorphp/rector-src/actions/runs/19303467511/job/55204152912?pr=7621#step:5:102

I am creating separate PR to fix it:

• [Deps] Bump to PHPStan ^2.1.32 and fix test #7622

[ruudk]

This makes the whole rule far less useful.

[calebdw]

This rule defeats the purpose of typing interfaces in your application (violates DIP). I've skipped this rule for this reason---I think it should be made optional and not part of the default type declaration set

[TomasVotruba]

Based on @ruudk and @calebdw , I think we should keep it as it is. Otherwise we have something in the middle nobody uses

[samsonasik]

Okay, I can focus on bumping phpstan 2.1.32 for now :)

• [Deps] Bump to PHPStan ^2.1.32 and fix test #7622

[ruudk]

I wasn't aware that this rule would be automatically enabled somewhere. Why is that?

[samsonasik]

@ruudk it enabled at part of list level:

rector-src/src/Config/Level/TypeDeclarationLevel.php

Line 136 in 1a19a34

NarrowObjectReturnTypeRector::class,

by PR:

• [TypeDeclaration] Register NarrowObjectReturnTypeRector to type declaration set #7615

[ruudk]

Ah that's also a bit of a YOLO mode isn't it? Whenever a new one is added?

[samsonasik]

yes, that's to test in the wild :)

the another one? removing type from union/nullable means dead code, so that's on dead code set

rector-src/src/Config/Level/DeadCodeLevel.php

Line 151 in 1a19a34

NarrowWideUnionReturnTypeRector::class,