Stochastic Differentially Private and Fair Learning