Add graceful shutdown support to listener

[coolguy1771]

Proposal

The listener currently exits immediately upon context cancellation, which can orphan sessions and interrupt in-flight message processing. I propose adding a configurable ShutdownTimeout to ensure the listener has a dedicated window to close the session and clean up resources before the process terminates.

Proposed Changes

1. Update Config: Add a ShutdownTimeout field (defaulting to 30s) to govern the cleanup phase.
2. Instrument Run: Use a defer block with a background context to ensure the Close signal is sent even after the main context is cancelled.

type Config struct {
    ScaleSetID      int
    MaxRunners      int
    ShutdownTimeout time.Duration // Added
    Logger          *slog.Logger
}

func (l *Listener) Run(ctx context.Context, scaler Scaler) error {
    timeout := l.config.ShutdownTimeout
    if timeout == 0 {
        timeout = 30 * time.Second
    }
    
    defer func() {
        // Use background context because the parent ctx is already cancelled
        cleanupCtx, cancel := context.WithTimeout(context.Background(), timeout)
        defer cancel()
        
        l.logger.Info("shutting down; cleaning up session")
        if err := l.client.Close(cleanupCtx); err != nil {
            l.logger.Error("session cleanup failed", "error", err)
        } else {
            l.logger.Info("session closed successfully")
        }
    }()
    
    // ... main loop ...
}

Key Benefits

• Reliability: Prevents "zombie" sessions from accumulating on the server side during deployments or restarts.
• Visibility: Adds logging to the shutdown sequence to help debug hung listeners.
• Safety: The use of a background context with a hard timeout ensures that the listener won't block indefinitely if the API is unresponsive during shutdown.