Differentially Private Stochastic Convex Optimization in (Non)-Euclidean Space Revisited