perf: Add MEMORY_TRADEOFFS and PERFORMANCE documentation

- Introduced MEMORY_TRADEOFFS.md to explain memory vs deduplication trade-offs in anagram generation.
- Added PERFORMANCE.md detailing optimizations for handling large volumes of anagram generation efficiently.
- Created USAGE.md for comprehensive usage instructions, including installation, basic commands, and advanced generation modes.
- Enhanced generator with streaming and batch processing capabilities for improved memory management.
- Implemented quick hashing for deduplication to reduce memory footprint.
- Updated main.rs to support new command-line arguments for streaming and batch modes.
- Added tests to ensure letter removal maintains minimum word length and to verify anagram sorting functionality.

tests/letter_removal_tests.rs

@@ -125,7 +125,7 @@ fn test_letter_removal_maintains_min_word_length() {
     // Should maintain at least 2 characters (word length - 1)
     for anagram in &anagrams {
         assert!(
-            anagram.text().len() >= 1,
+            !anagram.text().is_empty(),
             "Anagram '{}' is too short",
             anagram.text()
         );

tests/types_tests.rs

@@ -72,7 +72,7 @@ fn test_anagram_equality() {
 
 #[test]
 fn test_anagram_sorting() {
-    let mut anagrams = vec![
+    let mut anagrams = [
         Anagram::new("a".to_string(), PronouncabilityScore::new(50)),
         Anagram::new("b".to_string(), PronouncabilityScore::new(80)),
         Anagram::new("c".to_string(), PronouncabilityScore::new(65)),