Pipeline AutoMLStep notebook fixed

[CESARDELATORRE]

The issues fixed here are:

1. Notebook run failing. Error:

v2 driver import failed with exception: cannot import name 'setup_wrapper'. Falling back to v1 driver.
Traceback (most recent call last):
  File "setup_37a4b86d-1315-410d-92f5-d3ef02b27868.py", line 11, in <module>
    from azureml.train.automl._remote_script import setup_wrapper
ImportError: cannot import name 'setup_wrapper'
ImportError: cannot import name 'automl'

I think we don't need the RunConfiguration with the explicit Docker IMAGE and pip command. And this is what I think is causing the error above in versions missmatch related to the Docker image+pip. I simplified the notebook not to use the RunConfiguration and it works good for me now.

2. The AutoMLConfig is using outdated data approach by providing X and y instead of an AML Dataset and the name of the label column. That's why the user gets this warning:
   WARNING - The AutoMLConfig inputs you have specified will soon be deprecated. Please use the AutoMLConfig shown in our documentation: https://aka.ms/AutoMLConfig
   I also fixed that in my notebook by using an AML Dataset and a label name in the AutoMLConfig class.

3. Warning/Exception from the widget if the dataset doesn’t have a name.
   • If ‘dataset = Dataset.Tabular.from_delimited_files(example_data)’ doesn’t require a name/Id, why are we raising that exception below?
   • This is also a bug in the widget, so I created a bug in VSO about it: 615924

However, I created a WORKAROUND in the notebook by saving/registering the Dataset in the Workspace, so after that the dataset has a name and the widget doesn’t raise that exception/warning.

4. We should mention the "Compute Instance" instead of the "Notebook VM" that it mentions the notebook in its original version.

[CESARDELATORRE]

@anupsms When possible can you review this PR and merge if it looks good?

[lostmygithubaccount]

@CESARDELATORRE can we close this?

[CESARDELATORRE]

Closing since it's stale