Overcoming The Curse Of Dimensionality For Pdes: From Conventional Numerics And Nonlinear Monte Carlo To Deep Learning