Face parser: unify bounding box coordinates order

[edumucelli]

The face parser has two ways modus operandi:

1. Using one of the library's detector such as MTCNNFaceDetector or FaceAlignmentDetectoror
2. Giving a pre-computed bounding box.

However in (1) the coordinates order is y0, x0, y1, x1 (left, top, right, bottom) meanwhile on (2) it is y0, x0, y1, x1 (top, left, bottom,right).

That is very confusing when looking at the code and trying to use it without an internal face detector.

This PR reorders the coordinates and rename them so they are easily understandable.