mv_machine_learning: optimize FaceRecognition taskgroup

mv_machine_learning: optimize FaceRecognition taskgroup

[Issue type] code optimization

This patch optimizes the FaceRecognition task in the following way:

Avoiding reloading of model files: The model files after training
were loaded into the inference engine every time inference was
requested. This patch modifies this behaviour to make the
files load only if they were changed (i.e. model was retrained).

Change-Id: Ia7c588e22672b26329e513416f7f177bddbb1da6
Signed-off-by: Vibhav Aggarwal <v.aggarwal@samsung.com>