Collaborative Multi-agent Stochastic Linear Bandits