Add tests for unification of all types

[jackh726]

Originally mentioned in #531

It might be worth testing that specific types can unify with each other. In that PR, we specifically talked about trait objects with different ordering of auto traits. I imagine there are other types that we can specifically test for unification.

Some things off the top of my head that I can think about:

• Anything (including trait objects) with different ordering of where clauses
• Lifetimes in dyn
• Lifetimes in Raw and Ref types
• Lifetimes in closures, functions, etc.

c.c. @Areredify @zaharidichev

[zaharidichev]

@rustbot claim

[jackh726]

@rustbot release-assignment