Understanding Benign Overfitting in Nested Meta Learning