(yet) (Paper) Transformers Learn Higher-Order Optimization Methods for In-Context Learning, A Study with Linear Models