Benchmark detection performance

[ryanmasondavies]

We don't need to be super thorough about this, as we can assume Apple have done testing of their implementations themselves. However, we need to find out which API performs best in general out of the following:

• CIDetector
• AVCaptureMetadataOutput as mentioned by @D-Link13 in Very slow on older devices #3
• Any others? We could possibly check OpenCV if neither of the above are satisfactory, but that's a big dependency to include if the SDK already provides a good enough solution.

It's also possible both of these implementations are backed by the same algorithm, so performance might be identical. It would be interesting to find out!

[ryanmasondavies]

Some things to try:

• Benchmark performance of AVCaptureMetadataOutput for facial recognition running on the device
• Benchmark performance of CIDetector for facial recognition running on the device
• Does setting an image orientation affect performance / accuracy?
• Does resizing images (e.g making input images smaller) affect performance / accuracy?