Fix Identical type specification of constant types

[herndlm]

Closes phpstan/phpstan#7257

This was not working correctly because e.g. a ConstantArrayType cannot be safely used to strictly specify the other (left or right) type in the Identical comparison without further checks.

The problem with the linked issue was that it would have the same types on the left and right side and basically cancel out the types in a falsey scope (the right side of the || for e.g. $sorted leading to a NeverType), see https://phpstan.org/r/f560aac6-602f-45a0-8115-bb94f7142f79. Which explains why it works if only truthy scopes (via variable assignments) work around that.

[ondrejmirtes]

Thank you very much!

[ondrejmirtes]

There are bunch of code examples in current issues where the results changed because of this PR:

• What means "expects [some type here], *NEVER* given"? phpstan#6781
• Match arm comparison between *NEVER* and array{t1: bool, t2: bool} is always false. phpstan#6647
• Type changed to *NEVER* phpstan#5896
• False positive: Parameter x of function y expects z, *NEVER* given phpstan#5474
• Inside foreach, $value is not the same as $array[$key] phpstan#6000

Can you please go through them and see:

• If it's A) fixing a bug or B) introducing a new bug
• A): please send a regression test :)
• B): please reproduce the problem in a test and fix the test :)

Thank you very much!

[herndlm]

Looking good, do you have preferences in the regression tests? All in one PR/commit OK? Looks like I'll also have to create two tests for an issue or two.
I'll open them later today.

[ondrejmirtes]

Each closed issue should have a separate test case, meaning testBugXyz for rule test cases, or separate file in NodeScopeResolverTest :) They can be a single commit "Regression tests".