… a FnDef

Sometimes the parser needs to desugar a doc comment into `#[doc =
r"foo"]`. Currently it does this in a hacky way: by pushing a "fake" new
frame (one without a delimiter) onto the `TokenCursor` stack.

This commit changes things so that the token stream itself is modified
in place. The nice thing about this is that it means
`TokenCursorFrame::delim_sp` is now only `None` for the outermost frame.

The motivation here is to eliminate the `Option<(Delimiter,
DelimSpan)>`, which is `None` for the outermost token stream and `Some`
for all other token streams.

We are already treating the innermost frame specially -- this is the
`frame` vs `stack` distinction in `TokenCursor`. We can push that
further so that `frame` only contains the cursor, and `stack` elements
contain the delimiters for their children. When we are in the outermost
token stream `stack` is empty, so there are no stored delimiters, which
is what we want because the outermost token stream *has* no delimiters.

This change also shrinks `TokenCursor`, which shrinks `Parser` and
`LazyAttrTokenStreamImpl`, which is nice.

This makes it clear they return token trees, and makes for a nice
comparison against `TokenCursor` which returns tokens.

Autotrait bounds on dyn-safe trait methods

This PR is a successor to rust-lang#106604 implementing the approach encouraged by rust-lang#106604 (comment).

**I propose making it legal to use autotraits as trait bounds on the `Self` type of trait methods in a trait object.** rust-lang#51443 (comment) justifies why this use case is particularly important in the context of the async-trait crate.

```rust
#![feature(auto_traits)]
#![deny(where_clauses_object_safety)]

auto trait AutoTrait {}

trait MyTrait {
    fn f(&self) where Self: AutoTrait;
}

fn main() {
    let _: &dyn MyTrait;
}
```

Previously this would fail with:

```console
error: the trait `MyTrait` cannot be made into an object
 --> src/main.rs:7:8
  |
7 |     fn f(&self) where Self: AutoTrait;
  |        ^
  |
  = warning: this was previously accepted by the compiler but is being phased out; it will become a hard error in a future release!
  = note: for more information, see issue rust-lang#51443 <rust-lang#51443>
note: for a trait to be "object safe" it needs to allow building a vtable to allow the call to be resolvable dynamically; for more information visit <https://doc.rust-lang.org/reference/items/traits.html#object-safety>
 --> src/main.rs:7:8
  |
6 | trait MyTrait {
  |       ------- this trait cannot be made into an object...
7 |     fn f(&self) where Self: AutoTrait;
  |        ^ ...because method `f` references the `Self` type in its `where` clause
  = help: consider moving `f` to another trait
```

In order for this to be sound without hitting rust-lang#50781, **I further propose that we disallow handwritten autotrait impls that apply to trait objects.** Both of the following were previously allowed (_on nightly_) and no longer allowed in my proposal:

```rust
auto trait AutoTrait {}

trait MyTrait {}
impl AutoTrait for dyn MyTrait {}  // NOT ALLOWED

impl<T: ?Sized> AutoTrait for T {}  // NOT ALLOWED
```

(`impl<T> AutoTrait for T {}` remains allowed.)

After this change, traits with a default impl are implemented for a trait object **if and only if** the autotrait is one of the trait object's trait bounds (or a supertrait of a bound). In other words `dyn Trait + AutoTrait` always implements AutoTrait while `dyn Trait` never implements AutoTrait.

Fixes dtolnay/async-trait#228.

r? `@lcnr`

…=compiler-errors

Add candidates for DiscriminantKind builtin

Part of rust-lang#107379

…dex, r=lcnr

Emit warnings on unused parens in index expressions

Fixes: rust-lang#96606.

I am not sure what the best term for "index expression" is. Is there a better term we could use?

…=petrochenkov

Improve `TokenCursor`.

Some small improvements, for things that were bugging me.

Best reviewed one commit at a time.

r? ``@petrochenkov``

…=oli-obk

Don't cause a cycle when formatting query description that references a FnDef

When a function returns `-> _`, we use typeck to compute what the resulting type of the body _should_ be. If we call another query inside of typeck and hit a cycle error, we attempt to report the cycle error which requires us to compute all of the query descriptions for the stack.

However, if one of the queries in that cycle has a query description that references this function as a FnDef type, we'll cause a *second* cycle error from within the cycle error reporting code, since rendering a FnDef requires us to compute its signature. This causes an unwrap to ICE, since during the *second* cycle reporting code, we try to look for a job that isn't in the active jobs list.

We can avoid this by using `with_no_queries!` when computing these query descriptions.

Fixes rust-lang#107089

The only drawback is that the rendering of opaque types in cycles regresses a bit :| I'm open to alternate suggestions about how we may handle this...

…, r=Nilstrieb

Fix suggestion for coercing Option<&String> to Option<&str>

Fixes rust-lang#107604

This also makes the diagnostic `MachineApplicable`, and runs `rustfix` to check we're not producing incorrect code.

``@rustbot`` label +A-diagnostics