fix(nas_sync): Phase3 full 修复(STEP降片+merge PRAGMA)+全量完成2.02亿行

- sync_dbbardata.sh: STEP 2000万→200万/片(首片2GB export+scp跨公网挂)
- merge_increment.py: 加 PRAGMA cache=500MB+WAL+NORMAL+synchronous
  (replica 2亿行 UNIQUE索引数GB,默认2MB cache致索引IO merge>120s超时)
- 全量完成: replica 2.019亿行(=dbbardata全量日线+15min+ETF)
- 踩坑链: ssh reset中断→timeout杀ssh留NAS merge残留持锁→since落后数据已在库
- 教训: timeout杀ssh不杀远程python;merge大表必加PRAGMA;断点续传since.txt救命

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

scripts/nas_sync/merge_increment.py

@@ -30,6 +30,10 @@ def main():
 
     conn = sqlite3.connect(args.db)
     cur = conn.cursor()
+    # 大表(1.8亿行+)merge 提速: 默认2MB cache 致 UNIQUE索引(数GB)全磁盘IO, merge>120s超时
+    for _p in ("PRAGMA journal_mode=WAL", "PRAGMA synchronous=NORMAL",
+               "PRAGMA cache_size=-500000", "PRAGMA temp_store=MEMORY"):
+        cur.execute(_p)
     cur.execute(SCHEMA)
     cur.execute(
         "CREATE UNIQUE INDEX IF NOT EXISTS uq_dbbardata "

scripts/nas_sync/sync_dbbardata.sh

@@ -26,7 +26,7 @@ DB=$ROOT/data_backup/quant_trading.db
 STAGE=$ROOT/data_backup/_staging
 STATE=$ROOT/data_backup/since.txt
 LOG=$ROOT/data_backup/sync.log
-STEP=20000000  # 全量分片步长（行/片，~2GB/片）
+STEP=2000000  # 全量分片步长（行/片，~200MB/片；首版2000万/2GB致首片export+scp跨公网挂,降至此）
 
 mkdir -p "$STAGE"
 LOG_TS="$(date '+%F %T')"