Regret Analysis of the Finite-Horizon Gittins Index Strategy for Multi-Armed Bandits