Incentivized Exploration via Filtered Posterior Sampling