Better documentation and discoverability around Append and Transfomer Scope

As part of our public API, we support the ability to filter out parts of the estimator chain. The APIs that support this are not documented, nor do we provide any context in our Cookbooks. As we are doing an overall coverage of documentation/samples, I created this issue separately to ensure that it gets covered. Referencing from #3063, these are the two APIs that are used:

https://github.com/dotnet/machinelearning/blob/db4ecc0135baffc8930201a85fb3e9a101cdfa46/src/Microsoft.ML.Data/DataLoadSave/EstimatorChain.cs#L87

https://github.com/dotnet/machinelearning/blob/db4ecc0135baffc8930201a85fb3e9a101cdfa46/src/Microsoft.ML.Data/DataLoadSave/TransformerChain.cs#L145

It would also be very useful to have more information about why you would want to filter as well as what scenarios we are trying to address.