…iaskrgr

Rollup of 6 pull requests

Successful merges:

 - rust-lang#121543 (various clippy fixes)
 - rust-lang#122540 (Do not use `?`-induced skewing of type inference in the compiler)
 - rust-lang#122730 (Expose `ucred::peer_cred` on QNX targets to enable dist builds)
 - rust-lang#122732 (Remove redundant coroutine captures note)
 - rust-lang#122739 (Add "put" as a confusable for insert on hash map/set)
 - rust-lang#122748 (Reduce `pub` usage in `rustc_session`.)

r? `@ghost`
`@rustbot` modify labels: rollup

Rollup merge of rust-lang#122748 - nnethercote:rustc_session-pub, r=jackh726

Reduce `pub` usage in `rustc_session`.

In particular, almost none of the errors in `errors.rs` are used outside the crate.

r? `@jackh726`

Upgrades the Rust toolchain to `nightly-2024-03-21`. The relevant
changes in Rust are:
* rust-lang/rust#122480
* rust-lang/rust#122748
* rust-lang/cargo#12783

I wasn't confident that our regression testing could detect differences
in the file paths being generated with the new logic, so I added code
similar to the following just to check they were equivalent:
```rust
             let base_filename = tcx.output_filenames(()).output_path(OutputType::Object);
+            let binding = tcx.output_filenames(()).path(OutputType::Object);
+            let base_filename2 = binding.as_path();
+            assert_eq!(base_filename, base_filename2);
```
Note that this was done for each instance where we used `output_path`,
and completed regression testing with the previous toolchain version. I
also checked manually the instance where we generate a `.dot` graph for
debug purposes following the instructions
[here](https://model-checking.github.io/kani/cheat-sheets.html?highlight=dot#debug)
(a `libmain.dot` file was generated for the `main.rs` in one of my
projects).

---------

Co-authored-by: Celina G. Val <celinval@amazon.com>