fix: skip server cleanup on session loss to keep playback alive

Symptoms observed in logs — the TCP session dies, then cleanup fails:

  ERROR librespot_core::session  Connection to server closed.
  WARN  librespot_connect::spirc  unexpected shutdown
  ERROR librespot_core::session  Broken pipe (os error 32)
  ERROR librespot_core::session  Transport endpoint is not connected (os error 107)
  WARN  librespot  Spirc shut down unexpectedly

When SpircTask exits because session.is_invalid(), the post-loop
cleanup called handle_disconnect() (which sets play_status to Stopped
and tries to notify Spotify), delete_connect_state_request(), and
dealer().close(). All of these fail because the TCP connection is dead,
and setting play_status to Stopped needlessly kills the Player.

Now we detect session.is_invalid() and skip all server communication
in the post-loop cleanup. The Player runs in a separate thread and
continues playing from its audio buffer. main.rs will create a new
session and Spirc.

After fix — the "Broken pipe" and "Transport endpoint is not
connected" errors no longer appear after session loss. The Player
continues playing while the session reconnects (see next commit for
state restoration evidence).

Co-authored-by: Copilot <[email protected]>

Author: antoinecellerier

connect/src/spirc.rs

@@ -599,6 +599,14 @@ impl SpircTask {
             }
         }
 
+        if self.session.is_invalid() {
+            // Session TCP connection died — skip server communication that
+            // would fail anyway. The Player continues playing from its
+            // buffer independently; main.rs will create a new session.
+            warn!("session lost, skipping server cleanup");
+            return;
+        }
+
         if !self.shutdown && self.connect_state.is_active() {
             warn!("unexpected shutdown");
             if let Err(why) = self.handle_disconnect().await {