The curious case of TrainedWrapperEstimatorBase and friends

So, while I was doing another round of internalization, one thing that I internalized was this. (Basically, something to handle the shimming from the now internal `IDataTransform` interface to the new `ITransformer` interface, during the regrettable situations -- thankfully few -- where such a thing is still necessary. So this:

https://github.com/dotnet/machinelearning/blob/63a7654edaddaf0eea6af98ea762b01594d083ee/src/Microsoft.ML.Data/DataLoadSave/TransformWrapper.cs#L20

and this

https://github.com/dotnet/machinelearning/blob/63a7654edaddaf0eea6af98ea762b01594d083ee/src/Microsoft.ML.Data/DataLoadSave/TransformWrapper.cs#L151

Now, that's all fine, but after doing the necessary work it seemed that I could delete the estimator wrapper entirely, but then I see this very intriguing note.

https://github.com/dotnet/machinelearning/blob/63a7654edaddaf0eea6af98ea762b01594d083ee/src/Microsoft.ML.Data/DataView/RowToRowMapperTransform.cs#L50-L56

I do not understand what is going on here. This is not essential -- everything here is internal -- but it seems at least odd. This seems to indicate that this method and the class I wanted to delete have something to do with each other, but as far as I can tell they have nothing to do with each other whatsoever. But then why the comment?

Anyway, there's clearly something odd going on. This is absolutely not critical, but I wanted to register an issue about the oddness, since the code underlying it passes my understanding.