OpenFace 0.2.5

A number of changes to the interface and output of the model:
- Separation of landmark detection and action unit analysis, now the action unit module only requires location of facial landmarks to perform predictions, this allows for easy substitution of other facial landmark detection models, as long as they use the same 68 point markup
- Allowing for output of gaze angle and eye region landmarks
- Moving gaze analysis to a separate project
- Updating wiki to reflect these changes