Categorize community samples with structures similar to that of Onnx.Models

If the samples here are structured/categorized in the same way as that of the [Onnx.Models](https://github.com/onnx/models), then the community can better coordinate to address the ML.NET use cases that are still sparse or missing.

e.g Currently there are very few ML.NET samples on **Machine Comprehension,** except BERT.Net or **Machine Translation** (non ML.NET but based on Seq2SeqSharp)

By having the same structure as Onnx.Models, the ML.NET community can coordinate to focus on the missing categories like e.g Machine Comprehension or Machine Translation