mv_machine_learning: drop backbone model code dependency

mv_machine_learning: drop backbone model code dependency

[Issue type] code refactoring

Did code refactoring by dropping backbone model code dependency.

With this patch, we can change facenet to other without code modification.
Only what we have to do for the use of new backbone model is to update
each "value" attribute of two types of face_recognition.json file,
    "name"  : "FACE_RECOGNITION_INPUT_TENSOR_SHAPE",
    "type" : "array",
    "subtype" : "integer",
    "value" : [ 512, 1 ] <- input tensor shape of training model which is same
                            as output tensor shape of backbone model.

    "name"  : "FACENET_OUTPUT_TENSOR_NAME",
    "type"  : "string",
    "value" : "normalize/l2_normalize" <- output tensor name of backbone model.

Change-Id: I9e32c73e029d67a1c86e8b2c7b424cb09d614463
Signed-off-by: Inki Dae <inki.dae@samsung.com>