Direct model api

[rvashurin]

• Current WhiteboxModel#generate renamed to #pg_generate
• All calls to old generate renamed
• WhiteboxModel now acts like a proxy to it's underlying HF model - any unknown method calls will be redirected to it
• All calls to WhiteboxModel#device() replaced by #device, and old method removed - it will be responded by HF model directly
• New WhiteboxModel#generate method acts exactly like it's HF models method, and returns normal model output as it's first return value
• Second return value is result of #estimate_uncertainty call with the list of estimators specified as additional argument to #generate

Example of usage provided in temporary test_model_api.py script in the root.