Adjust design of unit tests with non-`()` type

[nikomatsakis]

Per this comment, we want to alter the design of unit tests with non-() type. The new design is:

• #[should_panic] can only be applied to tests with () return type and continues to have the same meaning.
• For non-should-panic tests, the return type T defines the error result using the Termination trait.

Achieving these changes should be relatively straight-forward. Basically the only thing we have to do is to make it an error to apply #[should_panic] unless the test has (). Currently we check that the test has the correct signature here:

rust/src/libsyntax/test.rs

Lines 330 to 355 in 9afed64

	fn has_test_signature(cx: &TestCtxt, i: &ast::Item) -> HasTestSignature {
	match i.node {
	ast::ItemKind::Fn(ref decl, _, _, _, ref generics, _) => {
	// If the termination trait is active, the compiler will check that the output
	// type implements the `Termination` trait as `libtest` enforces that.
	let output_matches = if cx.features.termination_trait_test {
	true
	} else {
	let no_output = match decl.output {
	ast::FunctionRetTy::Default(..) => true,
	ast::FunctionRetTy::Ty(ref t) if t.node == ast::TyKind::Tup(vec![]) => true,
	_ => false
	};
	no_output && !generics.is_parameterized()
	};
	if decl.inputs.is_empty() && output_matches {
	Yes
	} else {
	No
	}
	}
	_ => NotEvenAFunction,
	}
	}

In any case, we just have to extend that function to also check whether there is a should_panic attribute -- if so, we can report an error, sort of like this:

rust/src/libsyntax/test.rs

Line 365 in 9afed64

diag.span_err(i.span, "functions used as tests must have signature fn() -> ()");

[rcoh]

I'm going to try this right now 👍

[nikomatsakis]

Implemented in #49911