On Differentially Private Stochastic Convex Optimization with Heavy-tailed Data