The Limits of AI Explainability: An Algorithmic Information Theory Approach