Suggestion: store optimized hyperparameters in the zip file

**Is your feature request related to a problem? Please describe.**
During various AutoML runs I have generated multiple models with slightly different datasets. After using them in simulations I notice some perform better than others. I would like to re-train the model with slightly improved datasets. However, not all algorithms can be [retrained](https://docs.microsoft.com/en-us/dotnet/machine-learning/how-to-guides/retrain-model-ml-net). As I understand trees, for example, require me to create a new model. To create the new model I would like to use the hyperparameters from the earlier model.

**Describe the solution you'd like**
If the parameters are not stored at the moment, then I would suggest providing a way to store them to the zip file. At the moment, I did not notice any way to access them outside AutoML code. I can access them in Experiment.cs and PipelineSuggester.cs, but my current understanding is that I need customize the library myself to store them.

I think a good place to store hyperparameters would be inside TrainingInfo folder, in a new text file. Now it only includes version.txt

Ideally, ExperimentResult would include HyperParameters in the same way as the metrics.

I might be confused and just unable to find how to do it. If this is already possible, mention at end of https://docs.microsoft.com/en-us/dotnet/machine-learning/how-to-guides/retrain-model-ml-net would be nice.

**Additional context**
Getting the hyperparameters from AutoML experiments would be useful also for manual fine-tuning.

Model Builder does similar functionality already by creating the training file.