Distributed Value-Function Learning with Linear Convergence Rates