Customized data Featurization

**Is your feature request related to a problem? Please describe.**
I'm using Model Builder for text classification, but am unable to train with custom data preparation so Model builder uses English stop words

**Describe the solution you'd like**
I would like Model Builder to allow me to customize the data preparation pipeline (Example: FeaturizeText) before training.

**Describe alternatives you've considered**
I changed the code in *.training.cs that was generated after the first run of Model Builder and ran Model Builder again. but instead of using *.training.cs it generated a new one.