Naming rules of hyper-parameters and model help doc

[sneaxiy]

It seems that we have not unified the naming rules of hyper parameters.

• The TensorFlow models would use model.xxx, train.yyy , etc to specify the hyper parameters, while the hyper parameters of XGBoost models do not have suffix like model, train, etc.
• If the users define a new model using SQLFlow, and the constructor of the model is like def __init__(self, hp1, hp2). Should the users write WITH hp1=xxx, hp2=yyy or WITH model.hp1=xxx, model.hp2=yyy when writing SQLs in SQLFlow?
• Different models may have different settings, and we do not have a detailed doc for all settings of all official models.
• Furthermore, how about adding some grammar like help to get the setting doc of each model?

We need a unified naming rules of hyper parameters and something like help command for each model.