Benchmarking Representations for Speech, Music, and Acoustic Events