fix: 接入备份维护钩子并补齐 SQLX 缓存

- ProcessManager 新增 has_instance_process（备份启动守卫依赖）

- 删除/启动/编辑/迁移目录时接入备份维护钩子（lock_instance_maintenance + begin_instance_deletion + delete_instance_backups），remove_instance 拆出 remove_instance_files_and_state

- run 启动前检查活跃备份操作并加维护锁

- 目录迁移时搬移默认备份仓库并入队清理

- 更新 settings SELECT 的 SQLX 离线缓存（补 ui_font/backup_repository_path/mono_font）

- create_recorded_symlink 非 Windows 未用参数加 cfg_attr 豁免

packages/app-lib/src/api/instance/run.rs

@@ -75,6 +75,15 @@ async fn run_with_extra_launch_args_inner(
     extra_launch_args: Option<Vec<String>>,
     gc_intent: Option<GcLaunchIntent>,
 ) -> crate::Result<(ProcessMetadata, Option<GcLaunchReport>)> {
+    if crate::api::instance::has_active_instance_operation(instance_id) {
+        return Err(crate::ErrorKind::InputError(
+            "Wait for the active backup operation to finish before launching this instance"
+                .to_string(),
+        )
+        .into());
+    }
+    let _maintenance_guard =
+        crate::api::instance::lock_instance_maintenance(instance_id).await;
     let _hosted_guard =
         crate::pack::hosted::prepare_launch(instance_id, offline_mode).await?;
     let state = State::get().await?;