rustdoc: handle macro expansions in types

[el-ev]

• Closes ICE: rustdoc: unreachable #150154.
• Closes rustdoc: generate-macro-expansion produces incorrect expansion for macros in type position #150197.

This PR implements visit_ty in ExpandedCodeVisitor to correctly handle macro expansions in type positions (e.g., let _: foo!();). This also fixes a crash in rustdoc caused by the same issue.

Before:

After:

[rustbot]

r? @notriddle

rustbot has assigned @notriddle.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[fmease]

This is not the way to go about it. This should remain a panic since the callers should handle this pattern in a special way.

I'm pretty sure that this will lead to fn(: Ty) getting printed instead of fn(Ty). To fix this, callers should only print the pattern & the colon if the pattern isn't missing.

On my phone but I'm curious why we're only seeing this ICE now. This should be super simple to trigger via rustc -Zunpretty=normal or rustc -Zunpretty=expanded. Need to check.

Edit: Curiously, rustc correctly prints such signatures under both these modes, whether macro-generated or not, so I'm not sure yet how rustdoc is able to make ast-pretty reach this case.

[fmease]

Ah okay, so rustdoc manually calls pat_to_string inside macro_expansion. We should probably add a Panics section to pat_to_string instructing users to never print Missing since they should handle it on their own. And then we need to update rustdoc to do exactly that.

[rustbot]

Reminder, once the PR becomes ready for a review, use @rustbot ready.

[el-ev]

@rustbot ready

[el-ev]

Maybe this is not needed, as PatKind::Missing only appears in types. ccc3e42 should handle the two issues at once.