Clustering what Matters: Optimal Approximation for Clustering with Outliers