Minimization of Multi-Penalty Functionals by Alternating Iterative Thresholding and Optimal Parameter Choices