On the Convergence of Heterogeneous Reinforcement Learning Private Agents to Nash Equilibrium in a Macroeconomic Policy Game