No knee to be found when k=2 for metric='silhouette'

[Zhihao Zheng]

I noticed that for KElbowVisualizer(model, k=(2,12), metric='silhouette'), detecting knee and/or elbow will always fail when the largest silhouette score is at k=2, as k=1 is naturally not valid for silhouette calculation and the silhouette score curve will just directly fall, and there's no knee and elbow to be found. Yet the data do have an optimal number of clusters value, which is k=2.