perf: gate PartialEq derive to test builds only

The PartialEq derive was being compiled into release builds even though
it's only used in test code. This caused a 3.33% performance regression
in the du_human_balanced_tree benchmark due to increased binary size
affecting CPU cache efficiency.

Changes:
- stty.rs: Gate PartialEq derive on Flag<T> with #[cfg_attr(test, derive(PartialEq))]
- flags.rs: Gate PartialEq derive on AllFlags enum with #[cfg_attr(test, derive(PartialEq))]

This eliminates the performance regression while keeping all test code
functional and unchanged.

Author: naoNao89

src/uu/stty/src/flags.rs

@@ -27,7 +27,8 @@ use nix::sys::termios::{
     SpecialCharacterIndices as S,
 };
 
-#[derive(Debug, PartialEq)]
+#[derive(Debug)]
+#[cfg_attr(test, derive(PartialEq))]
 pub enum AllFlags<'a> {
     #[cfg(any(
         target_os = "freebsd",

src/uu/stty/src/stty.rs

@@ -63,7 +63,8 @@ const SANE_CONTROL_CHARS: [(S, u8); 12] = [
     (S::VDISCARD, 15), // ^O
 ];
 
-#[derive(Clone, Copy, Debug, PartialEq)]
+#[derive(Clone, Copy, Debug)]
+#[cfg_attr(test, derive(PartialEq))]
 pub struct Flag<T> {
     name: &'static str,
     #[expect(clippy::struct_field_names)]