refactor connected components algorithm to also support segmentation use case

[raybjork]

We don't support segmentation through our vision APIs yet but theres not a reason our library functions cant start to support this use case. I have an application I'm working on that requires this and it is not possible with existing vision support. This PR refactors the underlying algorithm we use for our detections to return clusters of pixels along with their bounding boxes.

I tested this functionality works by running a test that I don't think has a place in RDK but calls this function with the values in the config below

objectdetection.ColorDetectorConfig{
		SegmentSize:       20, 
		HueTolerance:      .9,
		SaturationCutoff:  .9,
		ValueCutoff:       .9,
		DetectColorString: "#FF660",
}

and applies them to this image

to get a mask of just the orange parts