test(maestro): fix borrow-after-partial-move in recover regression test

Capture sidecar_dir/fifo_dir/tmux/worktree_root before `drop(fx.manager)`
so the post-drop builder calls don't need to reborrow the partially-moved
fixture. Pure compile fix — same scenario as before.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

crates/maestro-daemon/tests/session_manager.rs

@@ -130,19 +130,28 @@ async fn recover_re_pipes_pane_so_output_reaches_subscribers() {
         .await
         .expect("create");
 
-    // Drop the original manager. The previous daemon's `pipe-pane` is still
-    // running against tmux, pointing at a FIFO no one reads now.
+    // Capture paths + tmux handle before dropping the original manager —
+    // `drop(fx.manager)` partially moves `fx` so we can't call its helpers
+    // afterwards.
+    let sidecar_dir = fx.sidecar_dir();
+    let fifo_dir = fx.fifo_dir();
+    let tmux = fx.tmux.tmux.clone();
+    let worktree_root = fx.worktree_root.path().to_path_buf();
+
+    // Drop the original manager. Its PtyBridge closes the FIFO's only
+    // reader, so tmux's `cat >> fifo` writer hits SIGPIPE and exits — but
+    // tmux's pipe-pane bookkeeping still thinks the pane is piped. That's
+    // exactly the stale state recover() has to overwrite.
     drop(fx.manager);
 
-    let sidecar = maestro_daemon::sidecar::SidecarStore::new(fx.sidecar_dir());
-    let worktree_mgr =
-        maestro_daemon::worktree::WorktreeManager::new(fx.worktree_root.path().to_path_buf());
+    let sidecar = maestro_daemon::sidecar::SidecarStore::new(sidecar_dir);
+    let worktree_mgr = maestro_daemon::worktree::WorktreeManager::new(worktree_root);
     let revived = maestro_daemon::SessionManager::new(
         "laptop".to_string(),
-        fx.tmux.tmux.clone(),
+        tmux,
         sidecar,
         worktree_mgr,
-        fx.fifo_dir(),
+        fifo_dir,
     );
     assert_eq!(revived.recover().await.expect("recover"), 1);