Currently, when detecting faces via Face Api, we are required to make multiple calls: 1 to detect faces, and 1 for each face to compare to a person group for matches.

Consider reducing this overhead, so that I can provide the PersonGroup to the Detect api, and have it return candidates in the same call.