`PartialOrd` doesn't work with new type

从 crates/farmer/ab-proof-of-space-gpu/src/shader/sort_buckets.rs 中的复现开始，重点查看第 17 行的 CustomPosition derive 和第 80 行的比较。将为 CustomPosition 生成的 PartialOrd 路径与直接的 u32 比较进行对比，然后检查对不需要的 Int8 capability 的处理；当 newtype 比较无需 Int8 即可编译时，即视为完成。

I have the following trivial new type:

#[derive(Debug, Clone, Copy, Ord, PartialOrd, Eq, PartialEq, From, Into)]
pub struct CustomPosition(u32);

When trying to compare two values of this type, I'm getting an error:

  error: `i8` type used without `OpCapability Int8`
       |
  note: used from within `<u32 as core::cmp::PartialOrd>::partial_cmp`
      --> /home/nazar-pc/.rustup/toolchains/nightly-2025-06-23-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/core/src/cmp.rs:1989:5
       |
  1989 |     ord_impl! { char usize u8 u16 u32 u64 u128 isize i8 i16 i32 i64 i128 }
       |     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  note: called by `<ab_proof_of_space_gpu::shader::sort_buckets::CustomPosition as core::cmp::PartialOrd>::partial_cmp`
      --> crates/farmer/ab-proof-of-space-gpu/src/shader/sort_buckets.rs:17:35
       |
  17   | #[derive(Debug, Clone, Copy, Ord, PartialOrd, Eq, PartialEq, From, Into)]
       |                                   ^^^^^^^^^^
  note: called by `<ab_proof_of_space_gpu::shader::sort_buckets::CustomPosition as core::cmp::PartialOrd>::le`
      --> /home/nazar-pc/.rustup/toolchains/nightly-2025-06-23-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/core/src/cmp.rs:1403:14
       |
  1403 |         self.partial_cmp(other).is_some_and(Ordering::is_le)
       |              ^^^^^^^^^^^^^^^^^^
  note: called by `ab_proof_of_space_gpu::shader::sort_buckets::perform_compare_swap`
      --> crates/farmer/ab-proof-of-space-gpu/src/shader/sort_buckets.rs:80:32
       |
  80   |     let (smaller, larger) = if a <= b { (a, b) } else { (b, a) };
       |                                ^^^^^^

At the same time, comparison of u32 directly works fine and doesn't require Int8 capability.

While this is likely related to discussion in https://github.com/Rust-GPU/rust-gpu/issues/307, I think it is distinct enough to be reported separately.