Deciphering Personalization: Towards Fine-Grained Explainability in Natural Language for Personalized Image Generation Models