Default RawOsError to i16 for 16-bit targets

[bushrat011899]

ACP: rust-lang/libs-team#755
Tracking issue: #154046 & #107792
Fixes #160541

Description

As noted by @tgross35, it's likely desirable for RawOsError to default to i16 for 16-bit platforms. Practically, this only affects AVR and MSP430. Since c_int is defined as:

crate::cfg_select! {
    any(target_arch = "avr", target_arch = "msp430") => {
        pub(super) type c_int = i16;
        pub(super) type c_uint = u16;
    }
    _ => {
        pub(super) type c_int = i32;
        pub(super) type c_uint = u32;
    }
}

This also ensures for all platforms (excluding UEFI), RawOsError is equivalent to c_int. Since neither sets of targets had an implementation of std, this hasn't yet been litigated. While RawOsError is unstable, it seems reasonable to choose a default more likely to align with the underlying OS (if any).

---

Notes

• No AI tooling of any kind was used during the creation of this PR.

[rustbot]

r? @Darksonn

rustbot has assigned @Darksonn.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

Why was this reviewer chosen?

The reviewer was selected based on:

• Owners of files modified in this PR: libs
• libs expanded to 12 candidates
• Random selection from 6 candidates

[Darksonn]

Sorry I'm at capacity for the next two weeks.

@rustbot reroll

[bushrat011899]

Clar I don't know why RustBot keeps giving you my PRs but hello again!

[lygstate]

well, the test already declare that.

View changes since the review

[bushrat011899]

Yeah tgross35's suggestion to use explicitly sized types is compared against the c_int assumption just so if a target is ever added that violates it, we know about it.

[tgross35]

Fwiw I was thinking to put it in a `const _: () = { /* ... */ } so the tests don't actually have to be run (I don't think we run coretests on these 16-bit platforms), but it doesn't matter much

[bushrat011899]

Yeah I'm not a huge fan of that particular const trick. The test as written is still compile-time evaluated, it'll just hopefully give a nicer message if/when it fails.