Raise when type parameters collide during sig to RBS translation

[Hashim1999164]

Summary

• RBS type variables shadow constants and class type members that share the same name, so a translated sig can fail to parse or drop a type error
• When a method type parameter collides with a constant path or class type member in the same signature, raise an error that asks the user to rename the type parameter

Closes #1012

Test plan

• bundle exec rake test

[jesse-shopify]

My concern with this is renaming a user's type name can lead to confusion if that name is surfaced later. I'm thinking a better strategy will be to raise an error suggesting the user rename to avoid the collision. That keeps the user in control and the names consistent.

[Hashim1999164]

@jesse-shopify hi, i understand your concern. Can you please take it up from here. I wint be available upcoming few days. Thanks

[Hashim1999164]

@jesse-shopify I switched this to raise instead of renaming. The error asks the user to rename the type parameter so the original names stay consistent and they stay in control.

[Hashim1999164]

@jesse-shopify yeah that makes sense. dropped the rename approach. it raises now and tells you to rename the type param yourself so names stay consistent.

[Hashim1999164]

@jesse-shopify CI was red on a Sorbet nilable thing for proc_returns / proc_bind. pushed a tiny fix that assigns them to locals before visit_type so the type narrows. should be green now

[Hashim1999164]

@jesse-shopify yeah this PR already raises instead of renaming, just had to rebase through a conflict on main. should be good now.

[Hashim1999164]

@jesse-shopify any update on this one? it's approved and rebased, just checking if anything else is needed before merge

[Hashim1999164]

@jesse-shopify any update on this? still approved on my side, just checking if its waiting on anything else to merge

[jesse-shopify]

This is still failing CI. I hope to look at it later today.

[Hashim1999164]

@jesse-shopify fixed the CI typecheck fail. ClassOf uses singular type_parameter in current RBI, so the walker was calling a method that doesnt exist. also merged latest main. should be green now.

[Hashim1999164]

@jesse-shopify one more thing: the Ruby workflow on the latest commit is waiting on maintainer approval for fork PRs (action_required). once that run is approved it should pick up the typecheck fix.