Expose max_det parameter in detect/segment mode

In detect/segment mode, YOLO limits the number of predicted objects to 300 by default. However, in some cases (newspapers) this is not enough.

We should expose the max_det parameter in the worker, and update the prediction to:

results = self.model(
    image, 
    device=self.device, 
    conf=self.confidence_threshold, 
+   max_det=self.max_detection
)