Incremental cluster-based retrieval using compressed cluster-skipping inverted files