Regret Bounds for Stochastic Shortest Path Problems with Linear Function Approximation