Phase 2: snapshot - window tree + screenshot via xcap

- Add xcap and image dependencies
- DesktopBackend trait with all 16 methods for future extensibility
- X11Backend with real snapshot() using xcap Window::all() and
  Monitor::all() for z-ordered window enumeration and screenshot
- Stub implementations for input/window management (phases 4-6)
- Wire X11Backend into DaemonState (now returns Result)
- Real snapshot handler replacing placeholder, updates ref map

src/daemon/mod.rs

@@ -46,7 +46,10 @@ async fn async_run() -> Result<()> {
         .context(format!("Failed to bind socket: {}", socket_path.display()))?;
 
     let session = std::env::var("DESKTOP_CTL_SESSION").unwrap_or_else(|_| "default".to_string());
-    let state = Arc::new(Mutex::new(DaemonState::new(session, socket_path.clone())));
+    let state = Arc::new(Mutex::new(
+        DaemonState::new(session, socket_path.clone())
+            .context("Failed to initialize daemon state")?
+    ));
 
     let shutdown = Arc::new(tokio::sync::Notify::new());
     let shutdown_clone = shutdown.clone();