This allows simplifying the call site and make_input by using a single
match instead of two levels of if's.

This allows re-entrant entering of the GlobalCtxt

We now only exit the GlobalCtxt when calling a callback and all the way
at the end when the GlobalCtxt is about to be destroyed.

There is no other query that may need to be called at that point anyway.

Several custom drivers are incorrectly calling queries.global_ctxt()
from inside of it, which causes some driver code to be skipped. As such
I would like to either remove it in the future or if custom drivers
still need it, change it to accept an &rustc_ast::Crate instead.

A used function with no mappings has historically indicated a bug, but that
will no longer be the case after moving some fallible span-processing steps
into codegen.

This will avoid confusion with actual `Span` spans.

…t are not WF

It's okay though b/c these are duplicated diagnostics.

…_round, r=cjgillot

Some more refactorings towards removing driver queries

Follow up to rust-lang#127184

## Custom driver breaking change

The `after_analysis` callback is changed to accept `TyCtxt` instead of `Queries`. The only safe query in `Queries` to call at this point is `global_ctxt()` which allows you to enter the `TyCtxt` either way. To fix your custom driver, replace the `queries: &'tcx Queries<'tcx>` argument with `tcx: TyCtxt<'tcx>` and remove your `queries.global_ctxt().unwrap().enter(|tcx| { ... })` call and only keep the contents of the closure.

## Custom driver deprecation

The `after_crate_root_parsing` callback is now deprecated. Several custom drivers are incorrectly calling `queries.global_ctxt()` from inside of it, which causes some driver code to be skipped. As such I would like to either remove it in the future or if custom drivers still need it, change it to accept an `&rustc_ast::Crate` instead.

coverage: Store coverage source regions as `Span` until codegen

Historically, coverage spans were converted into line/column coordinates during the MIR instrumentation pass.

This PR moves that conversion step into codegen, so that coverage spans spend most of their time stored as `Span` instead.

In addition to being conceptually nicer, this also reduces the size of coverage mappings in MIR, because `Span` is smaller than 4x u32.

---

There should be no changes to coverage output.

…=joboet

Add missing code examples on `LocalKey`

r? ``@Amanieu``

…ve-never, r=lcnr

Structurally resolve before checking `!` in HIR typeck

Some more missing structural resolves in HIR typeck :>

r? lcnr

…ve-cat-proj, r=lcnr

Structurally resolve before matching on type of projection

Another missing structural resolve in closure upvar analysis. I think it's better to place the normalization here rather than trying to guarantee that all types returned by the expr use visitor are structurally normalized, which I don't think we do now. Thoughts?

r? lcnr