Optimize distance calculation algorithm

[MB-Finski]

I know I'm a bit late to the party what with the ex-apps and all but here's an improved distance calculation algorithm for face recognition clustering:

Since, the calculation of vector-to-vector distances is done almost to the tune of face detections squared (likely millions of times for each clustering run), even small improvements here should translate to considerable improvements in the overall performance.

Benchmark results for 1000000 distance calculations:
------------------------------------------------------------
Original Method (foreach & ** 2):       7.921905 seconds
Optimized Method (for loop & * 2):      3.993280 seconds
------------------------------------------------------------

[marcelklehr]

Woop woop, awesome :) Thank you!
Looking forward to more PRs from you 😀