NI: don't use constraints with StubType when determining type variable readiness

[jsjeon]

#KT-42591 Fixed

[jsjeon]

Hi @KirpichenkovPavel

This draft attempts to postpone adding (improper) type constraints for stub types, which is part of builder inference, and is able to unmute a bb test as well as identify incorrectly resolved references. However, it has a couple regressions: https://teamcity.jetbrains.com/buildConfiguration/Kotlin_KotlinForGoogle_AggregateBranch/3266636?buildTab=tests&status=failed but I wonder if it's worth digging into further, i.e., this is the right direction to resolve KT-42591. Your early input would be really appreciated, thanks!

Cc'ed @demiurg906 as it is about FIR builder inference.

[demiurg906]

@PetukhovVictor can you please take a look? There are changes in common part of inferences between two frontends