Better way to create and inform users the support of sampling API on each model

[edwinnglabs]

Is your feature request related to a problem? Please describe. Right now it requires manually created functions to create Forecaster and the conditions on each supported estimators

Describe the solution you'd like We can do some organization with Enums and factory style of functions to produce Forecaster with different supported estimators.

reference to gluonTS; core parameters in estimator and model