The Sweet Spot between Inverted Indices and Metric-Space Indexing for Top-K-List Similarity Search