운영 서버에서 DB index 적용 작업 진행

- 01~07 작업 문서를 운영 측정값 기준으로 정리 (문서의 DB 비밀번호 삭제)
- app_highest_record 중복 8,795행 점수 기준 정리, 인덱스 5개 + UNIQUE 키 2개 추가
- 02·04 공통 측정 스크립트와 측정 결과, comparison.txt 추가
- 서버 처리 시간 18~292배 단축, 결과 행 수 동일
- 운영 데이터 백업과 권한(비밀번호 해시) 파일은 커밋에서 제외

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

doc/plan/db/260915-improve-production/result/dedupe_result.txt

@@ -0,0 +1,47 @@
+--------------
+DELETE AHR FROM app_highest_record AHR
+JOIN (
+  SELECT AppHighestRecordID FROM (
+    SELECT AppHighestRecordID,
+           ROW_NUMBER() OVER (
+             PARTITION BY MaestroID, PlayerID, AppID
+             ORDER BY CASE WHEN AppID = 105 THEN HighestRecord ELSE -HighestRecord END,
+                      AppHighestRecordID DESC
+           ) AS rn
+    FROM app_highest_record
+  ) ranked
+  WHERE rn > 1
+) dup ON dup.AppHighestRecordID = AHR.AppHighestRecordID
+--------------
+
+Query OK, 8795 rows affected (0.942 sec)
+
+--------------
+SELECT 'app_highest_record' AS table_name,
+       COUNT(*) AS total_rows,
+       COUNT(DISTINCT MaestroID, PlayerID, AppID) AS unique_combos
+FROM app_highest_record
+--------------
+
++--------------------+------------+---------------+
+| table_name         | total_rows | unique_combos |
++--------------------+------------+---------------+
+| app_highest_record |     220648 |        220648 |
++--------------------+------------+---------------+
+1 row in set (0.190 sec)
+
+--------------
+SELECT 'typing_exam_highest_record' AS table_name,
+       COUNT(*) AS total_rows,
+       COUNT(DISTINCT MaestroID, PlayerID, WritingID) AS unique_combos
+FROM typing_exam_highest_record
+--------------
+
++----------------------------+------------+---------------+
+| table_name                 | total_rows | unique_combos |
++----------------------------+------------+---------------+
+| typing_exam_highest_record |      20413 |         20413 |
++----------------------------+------------+---------------+
+1 row in set (0.024 sec)
+
+Bye