The Importance Of Eigenstructure In High-Dimensional Statistics: Examples From Overparameterized Machine Learning And Graphical Models