Expediting On-Device LLM Personalization via Explainable Model Selection