mv_machine_learning: use DEFAULT_MODEL_NAME for landmark detection task group

mv_machine_learning: use DEFAULT_MODEL_NAME for landmark detection task group

[Issue type] : new feature

Use the meta file for landmark detection tasks to set the default model

With this patch, a proper concrete class object will be created
at the constructor of FacialLandmarkAdapter and PoseLandmarkAdapter classes
according to the default model names which are parsed from each meta file,
facial_landmark.json and pose_landmark.json files.

And also this patch calls the member function, create, to create the concrete
class object corresponding to the default model name instead of creating
a default concrete class object directly.

Finally, we can create user-desired concrete class object without rebuilding
by simply modifying the meta files.

Change-Id: I4ef045fe1ae24177ae4ebc7a3f48ae6e136c01dc
Signed-off-by: Inki Dae <inki.dae@samsung.com>