On Computational and Statistical Challenges of Solving Nonconvex Minimax Optimization Problems