Improve ONNX Object Detection Flow

[luisquintanilla]

Refactored sample to improve flow / readability

List of Changes

1. Renamed YoloWinMlParser to YoloOutputParser.
2. Fixed namespaces on classes within subdirectories to match their location in the project structure.
3. Changed return type of LoadModel method to ITransformer
4. Removed CreateEmptyDataView since it is only used once and replaced it with _mlContext.Data.LoadFromEnumerable(new List<ImageNetData>()); inside of LoadModel method.
5. Changed PredictDataUsingModel method signature to decouple post-processing steps.
6. Use Transform method instead of PredictionEngine to score data using the model in PredictDataUsingModel method.
7. Updated Score method to reflect changes in other helper methods.
8. Removed GetImagesData method. Moved logic to ImageNetData class as part of ReadFromFile method.
9. Moved DrawBoundingBox method to Program class
10. Expanded logic in Main method of Program class and broke down into steps (1) load data, (2) score data with model (3) post-process data (4) draw bounding boxes

[natke]

This could use a comment to explain why we create an empty data view

[CESARDELATORRE]

Tested and reviewed. Thanks! 👍