mv_machine_learning: use template type for outgoing_queue

mv_machine_learning: use template type for outgoing_queue

[Issue type] : code cleanup

Use template type for _outgoing_queue. As for asynchronous inference,
we need to allow various structures to be pushed to outgoing queue
because each task group can be used for different purpose and this means
that various result structures are needed.

This is just a step for introducing asynchronous inference manager.

Change-Id: I5d7fbe59da3c95e59366f10f8232eed42a6a5e60
Signed-off-by: Inki Dae <inki.dae@samsung.com>