SSAMBA: Self-Supervised Audio Representation Learning with Mamba State Space Model