CompileTests: don't help the compiler quite as much

[mpollmeier]

We want to assert that the return type is derived correctly. The way
we tested this (before this PR) gave the compiler more information
than it strictly needs, i.e. now we make the compiler's job harder, to
ensure it's always derived correctly.

[bbrehm]

Sorry for not spotting that earlier, but:

Why are you using intersection types here?

CallBase & StaticType[IsStaticT] is cute, but I'd feel better with the conservative type StaticCallBase = CallBase with StaticType[IsStaticT].

[mpollmeier]

¯_(ツ)_/¯ I used intersection types because that's what the compiler derives them to.
I like the syntax and our use case nicely aligns with their purpose as of https://docs.scala-lang.org/scala3/book/types-intersection.html

That's just my personal preference though, I'm happy to change it, but can you add more arguments than just 'feeling better'? :)

[bbrehm]

My argument is basically: Scala 3 added fancy new union types that exist in scala but not the jvm. The intersection types use a similar syntax. If you encounter that as a reader, then you're forced to take a detour into language/compiler details to understand that nothing is happening. The old Foo with Bar syntax is well-known and doesn't force a hickup.

But upon reconsideration, meh, we can keep it that way with the intersection type -- afaict nothing fancy is happening here.

[mpollmeier]

👍🏻 thanks for clarifying, will leave as is, and if you change your mind we can change this to old school composition