sort: pre-compute ICU collation sort keys for fast locale sorting

Add a fast_locale_collation path that pre-computes ICU sort keys during
line parsing (O(n)), then uses cheap byte-array comparison during
sorting (O(n log n)). This replaces per-comparison ICU compare_utf8
calls for the common case of plain `sort` with a UTF-8 locale.

With 1M lines and LC_ALL=en_US.UTF-8, this is ~2.6x faster than
GNU sort (314ms vs 822ms).

src/uu/sort/src/chunks.rs

@@ -75,7 +75,7 @@ impl LineData<'_> {
 impl Chunk {
     /// Destroy this chunk and return its components to be reused.
     pub fn recycle(mut self) -> RecycledChunk {
-        let recycled_contents = self.with_dependent_mut(|_, contents| {
+        let mut recycled_contents = self.with_dependent_mut(|_, contents| {
             contents.lines.clear();
             contents.line_data.selections.clear();
             contents.line_data.num_infos.clear();
@@ -100,7 +100,7 @@ impl Chunk {
                     &mut contents.line_data.selections,
                 ))
             };
-            (
+            RecycledChunk {
                 lines,
                 selections,
                 num_infos: std::mem::take(&mut contents.line_data.num_infos),

src/uu/sort/src/sort.rs

@@ -48,7 +48,7 @@ use uucore::error::{FromIo, strip_errno};
 use uucore::error::{UError, UResult, USimpleError, UUsageError};
 use uucore::extendedbigdecimal::ExtendedBigDecimal;
 #[cfg(feature = "i18n-collator")]
-use uucore::i18n::collator::locale_cmp;
+use uucore::i18n::collator::{compute_sort_key_utf8, locale_cmp};
 use uucore::i18n::decimal::locale_decimal_separator;
 use uucore::line_ending::LineEnding;
 use uucore::parser::num_parser::{ExtendedParser, ExtendedParserError};
@@ -324,6 +324,7 @@ struct Precomputed {
     floats_per_line: usize,
     selections_per_line: usize,
     fast_lexicographic: bool,
+    fast_locale_collation: bool,
     fast_ascii_insensitive: bool,
     tokenize_blank_thousands_sep: bool,
     tokenize_allow_unit_after_blank: bool,
@@ -387,6 +388,8 @@ impl GlobalSettings {
 
         self.precomputed.fast_lexicographic =
             !disable_fast_lexicographic && self.can_use_fast_lexicographic();
+        self.precomputed.fast_locale_collation =
+            disable_fast_lexicographic && self.can_use_fast_lexicographic();
         self.precomputed.fast_ascii_insensitive = self.can_use_fast_ascii_insensitive();
     }
 
@@ -632,6 +635,15 @@ impl<'a> Line<'a> {
         token_buffer: &mut Vec<Field>,
         settings: &GlobalSettings,
     ) -> Self {
+        #[cfg(feature = "i18n-collator")]
+        if settings.precomputed.fast_locale_collation {
+            compute_sort_key_utf8(line, &mut line_data.collation_key_buffer);
+            line_data
+                .collation_key_ends
+                .push(line_data.collation_key_buffer.len());
+            return Self { line, index };
+        }
+
         let needs_line_data = settings.precomputed.needs_tokens
             || settings.precomputed.selections_per_line > 0
             || settings.precomputed.num_infos_per_line > 0
@@ -2614,6 +2626,18 @@ fn compare_by<'a>(
         };
     }
 
+    #[cfg(feature = "i18n-collator")]
+    if global_settings.precomputed.fast_locale_collation {
+        let a_key = a_line_data.collation_key(a.index);
+        let b_key = b_line_data.collation_key(b.index);
+        let cmp = a_key.cmp(b_key);
+        return if global_settings.reverse {
+            cmp.reverse()
+        } else {
+            cmp
+        };
+    }
+
     if global_settings.precomputed.fast_ascii_insensitive {
         let cmp = ascii_case_insensitive_cmp(a.line, b.line);
         if cmp != Ordering::Equal || a.line == b.line {