[ctz]

The goal of this PR is to arrange that the compiler complains about GHSA-rc3f-4j82-cmrg in a useful way, and then spread that pattern as far as possible.

With reordering the commits in this PR, this is achieved:

error[E0597]: `server_name` does not live long enough
   --> librustls/src/ffi.rs:470:46
    |
470 |               unsafe { $crate::ffi::bound_cstr(&$var) }
    |                                                ^^^^^ borrowed value does not live long enough
    |
   ::: librustls/src/client.rs:886:9
    |
886 |           server_name: *const c_char,
    |           ----------- binding `server_name` declared here
...
890 |           let server_name = try_cstr!(server_name);
    |                             ---------------------- in this macro invocation
...
901 | /                 ClientConnection::new_with_alpn(
902 | |                     try_clone_arc!(config),
903 | |                     server_name,
904 | |                     alpn_protocols,
905 | |                 )
    | |_________________- argument requires that `server_name` is borrowed for `'static`
...
910 |       }
    |        - `server_name` dropped here while still borrowed
    |
    = note: this error originates in the macro `try_cstr` (in Nightly builds, run with -Z macro-backtrace for more info)

[cpu]

Thanks! This looks good to me.

I tacked on two commits at the end to try and get some targetted regression coverage. Those two commits applied to main without the rest of the rework fail in CI as expected.

Feel free to drop/rework them if you have opinions.

[cpu]

#699 should fix the macOS CI failure.

[cpu]

documentation / Generate API documentation (pull_request) - Failing after 1m

I think you need to run the docgen too (sorry, lots of manual steps in this repo).