Best/recommended practices, tools and workflows for AML

I find the current development of AzureML very confusing.

- There appears to be a transition from Python SDK v1 to v2 (confirmed)
- There appears to be a transition from web service endpoints to real-time / batch endpoints (uncomfirmed)
- There appears to be a transition from AzureML tools to using MLflow instead (unconfirmed)
- There appears to be a transition from using a score.py file to using MLflow model wrappers (unconfirmed)

If that's not confusing enough, this repo includes a folder called "python-sdk" that uses SDK v1 and another folder called "sdk" which uses SDK v2. To top things off there's a folder with notebooks that use both versions of the SDK.

Can you please:

- Restructure the folders in a meaningful way
- Provide recommendations on which functionalities to use and how

Start by comparing the python-sdk and sdk folders and reviewing the notebooks that use both SDK versions. Define a meaningful organization and document recommendations for SDK versions, endpoint types, tooling, and model packaging; done means the folder structure and guidance consistently answer which workflows to use.

Written by the indexing model from the issue text.