fix(deploy): deploy/ 自己没跟着上机，改它等于没改

重启跑的是 NAS 上那份 stop.sh / start.sh，而 push.sh 只同步 backend/ 和
client/build/。结果就是：修好 stop.sh 的 pgrep、提交、部署、报成功，远端
照样跑着旧脚本——上一个提交正是这么被架空的。

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

deploy/push.sh

@@ -62,6 +62,14 @@ $TAR --exclude .venv --exclude .env --exclude __pycache__ \
     --exclude '*.db' --exclude tests --exclude .pytest_cache \
     -C "$REPO/backend" . | sh_ "tar xzf - -C '$APP/backend'"
 
+# The restart below runs the NAS's own copies of stop.sh / start.sh, so they
+# have to travel with the code — otherwise a fix to them lands in git, the
+# deploy reports success, and the far side keeps running the old ones. That is
+# exactly how the broken pgrep guards survived a deploy that was meant to fix
+# them.
+echo "==> deploy scripts"
+$TAR -C "$REPO/deploy" . | sh_ "tar xzf - -C '$APP/deploy'"
+
 echo "==> static (cleared first, so stale JS chunks do not pile up)"
 if [ ! -f "$REPO/client/build/index.html" ]; then
   echo "client/build is missing — run 'npm run build' first" >&2