Next MSRV bump

Trixie released with rustc 1.85, so I expect a lot of crates will be increasing their MSRV in the next few weeks. I'd like to upgrade only as needed, rough list of what might be nice to have:

Baseline: 1.63

• 1.64: core::ffi::c_* (definitely need this)
• 1.64: workspace-level lints (meh)
• 1.65: cast_mut (this would be nice)

Baseline: 1.65

• 1.67: varargs for non-C extern ABIs (no known applications)
• 1.69: derive on packed structs that don't implement Copy (not overly useful for us, everything is copy)
• 1.75: byte_add (this would be nice)
• 1.75: const_maybe_uninit_zeroed (this would be nice)
• 1.77: offset_of (just lets us remove a polyfill)
• 1.77: C string literals (nothing we can't work around, not sure we even use this now)
• 1.78: cfg(target_abi = "...") https://github.com/rust-lang/libc/pull/5065#issuecomment-4861400142
• 1.80: size_of, align_of in prelude (nice to have but we have our own prelude)
• 1.80: variadics without a named parameter (no known applications)
• 1.81: #[expect(...)] (want this, but it's not pressing)
• 1.82: unsafe extern blocks (neutral)
• 1.82: &raw (few small applications, addr_of! works fine)
• 1.82: nested access in offset_of! (no uses)
• 1.84: ptr::without_provenance (some 0x1234 as *const T casts)
• 1.85: ptr::fn_addr_eq (we don't really need this as it is just for trait impls)
• 1.85: 2024 edition
• 1.92: &raw [mut|const] to union fields is now safe

No file or test is named. Start by reviewing the workspace's current MSRV and checking the listed Rust features against the crates that use them. Done means selecting and documenting a justified MSRV increase only where the project needs it, including whether the 2024 edition or later features are required.