Kmeans (sklearn vs FAISS), finding n_clusters via inertia/silhouette, Agglomorative, DBSCAN/HDBSCAN