An LP-based Sampling Policy for Multi-Armed Bandits with Side-Observations and S • Libertify