An optimal Gauss–Markov approximation for a process with stochastic drift and applications