mv_machine_learning: code refactoring to getOutput function

mv_machine_learning: code refactoring to getOutput function

[Issue type] : code refactoring

Do code refactoring to getOutput function of the concrete class,
ObjectDetection, by extracting some portion of existing getOutput function code
and moving it to a new interface function, getOutputCache.

With original version of the getOutput function, it handled the specific case
that one more get-result API are called, which in turn, it made the getOutput
function to be complicated.

Therefore, this patch drops the specific case from the getOutput function
by introducing getOutputCache interface which always returns internal
result object, _current_result instead of calling the result member function
of each concrete class which decodes raw output tensor data after the completion
of the inference.

Change-Id: I4aacb8fc55be3f63a983182b919291dcb927b0c9
Signed-off-by: Inki Dae <inki.dae@samsung.com>