Langevin dynamics based algorithm e-THεO POULA for stochastic optimization problems with discontinuous stochastic gradient