$\lambda$-Regularized A-Optimal Design and its Approximation by $\lambda$-Regularized Proportional Volume Sampling