[BUG] Silhouette Score Error

[rsshah1993]

Describe the bug
It appears that the calculation for the Silhouette scoring is inconsistent especially with singleton clusters. See below:

Steps/Code to reproduce bug

>>> from cuml.metrics.cluster import silhouette_samples
>>> import numpy as np
>>> vecs = np.array([[0.0, 0.0, 0.0], [1.0, 1.0, 1.0], [2.0, 2.0, 2.0], [10.0, 10.0, 10.0]])
>>> labels = np.array([0, 0, 1, 3])
>>> silhouette_samples(X=vecs, labels=labels)
array([0.5, 0. , 0. , 1. ])

It looks like for vecs[3] (which is a singleton cluster) produces a score of 0 while vecs[4] produces a score of 1. Is this the expected behavior? Another example:

>>> vecs = np.array([[0.0, 0.0, 0.0], [1.0, 1.0, 1.0], [10.0, 10.0, 10.0]])
>>> labels = np.array([1, 1, 3])
>>> silhouette_samples(X=vecs, labels=labels)
array([1., 1., 1.])

Expected behavior
I believe example one should return: [0.5, 0. , 0. , 0. ] and example two should return: [0.9, 0.888888888888889, 0.0]. It looks like sklearn assigns singleton clusters a score of 0 where as matlab assigns singleton clusters a score of 1 but still not sure if that explains either example.

Environment details:

• Environment location: Docker
• Linux Distro/Architecture: ubuntu18.04
• GPU Model/Driver: T4/455.32.00
• CUDA: 11.1
• Method of cuDF & cuML install: conda install -c rapidsai-nightly -c nvidia -c conda-forge -c defaults rapids=0.19 python=3.7 cudatoolkit=10.2