Fundamental Entropic Laws and $\mathcal{L}_p$ Limitations of Feedback Systems: Implications for Machine-Learning-in-the-Loop Control