numfmt: fix zero-padding placement for negative numbers (#11694)

* numfmt: fix zero-padding placement before sign for negative numbers

When using zero-padded format (e.g. %018.2f), the sign character is now
placed before the padding zeros, matching C printf behavior.

Fixes #11664

* numfmt: simplify sign handling with combined strip_prefix

src/uu/numfmt/src/format.rs

@@ -581,7 +581,12 @@ fn format_string(
     let padded_number = match padding {
         0 => number_with_suffix,
         p if p > 0 && options.format.zero_padding => {
-            let zero_padded = pad_string(&number_with_suffix, p as usize, '0', true);
+            let zero_padded = if let Some(unsigned) = number_with_suffix.strip_prefix(['-', '+']) {
+                let sign = &number_with_suffix[..1];
+                format!("{sign}{}", pad_string(unsigned, p as usize - 1, '0', true))
+            } else {
+                pad_string(&number_with_suffix, p as usize, '0', true)
+            };
 
             match implicit_padding.unwrap_or(options.padding) {
                 0 => zero_padded,

tests/by-util/test_numfmt.rs

@@ -1428,7 +1428,6 @@ fn test_from_unit_fractional_precision_issue_11663() {
 // Zero-padded `--format` places padding zeros before the sign for negative
 // numbers; GNU (and C printf) puts the sign first.
 #[test]
-#[ignore = "GNU compat: see uutils/coreutils#11664"]
 fn test_zero_pad_sign_order_issue_11664() {
     new_ucmd!()
         .args(&["--from=none", "--format=%018.2f", "--", "-9869647"])