fix: restart backend under systemd and use .sproutclaw agent dir

Restart via systemctl when running as a service, set SPROUTCLAW_CODING_AGENT_DIR
for the RPC subprocess, and add a frontend poll timeout for failed restarts.

Co-authored-by: Cursor <cursoragent@cursor.com>

frontend/src/routes/SettingsPage.tsx

@@ -100,7 +100,7 @@ function extensionCategory(extension: ExtensionInfo): "local" | "npm" {
 	}
 	if (extension.source?.startsWith("npm:")) return "npm";
 	const pathHint = `${extension.path || ""} ${extension.location || ""}`.replace(/\\/g, "/");
-	if (pathHint.includes("/.pi/agent/npm/node_modules/")) return "npm";
+	if (pathHint.includes("/.sproutclaw/agent/npm/node_modules/")) return "npm";
 	return "local";
 }
 
@@ -701,8 +701,13 @@ export function SettingsPage() {
 		} catch {
 			// 后端重启时连接断开属正常现象，忽略错误
 		}
-		// 轮询直到后端恢复
+		const deadline = Date.now() + 60_000;
 		const poll = async () => {
+			if (Date.now() > deadline) {
+				setRestarting(false);
+				setStatus("后端重启超时，请检查 sproutclaw-web 服务状态", "error");
+				return;
+			}
 			try {
 				await settingsApi.fetchSettings();
 				setStatus("后端已重启，正在刷新...", "ok");