Sparks of Large Audio Models: A Survey and Outlook