Always require semicolon or newline before method body

[FnControlOption]

As per #6191 (comment)

[straight-shoota]

I believe the same problem applies to many more uses of skip_token_end. So it's probably better to look at other uses as well. And extract the common behaviour to a dedicated method (consume_token_end maybe).

[straight-shoota]

Thanks for the enhancements and the very clear commit history 🙇

Specs are broken.

[straight-shoota]

Thanks for the many recent additions. I understand you have incorporated an implementation of the changes discussed in #11209?

To me that feels a bit much for this PR, though. Statement end before method body should be considered a separate feature from what comes after the name of a type def.

If you don't mind, I can take care of extracting this concern to a different PR.

[FnControlOption]

Opened #13375

As for this PR, I think I need to go back to the drawing board. def f(x) puts x end is both valid Ruby and Crystal, and requiring a semicolon in this case seems unnecessarily restrictive. I think a semicolon/newline should only be required if the method header (a) has both no parameters and no parentheses; or (b) has anything after the parentheses (i.e. an explicit return type and/or any generic type parameters).

[straight-shoota]

I created #13386 to discuss the intended behaviour.