Understanding Memorization in Generative Models via Sharpness in Probability Landscapes