Use boto3 to get credentials.

[ellisonbg]

I did some research about the best way to get the AWS credentials and it looks like boto3 handles all of the most common cases in a standard manner that should work well in most cases. This simplifies the logic while fixing some of the cases that weren't working (such as pulling credentials from the EC2 instance metadata).

Warning: I wrote in code in the GitHub web editor so I didn't run it locally to verify it works. Please test locally before merging :-)

[github-actions]

👈 Launch a Binder on branch ellisonbg/jupyter-drives/s3-credentials

[DenisaCG]

Thank you @ellisonbg!

I tested it locally and it works well. I re-added some of the previously deleted logic, such that we can keep the extension open to other providers.

Now the extension will automatically check jupyter_notebook_config.py for credentials, if it is empty, it will try to extract the S3 credentials through boto and, if that is not possible, check the environment variables (other providers can make use of this). This way we can enable a seamless credentials extraction for S3, but also keep the logic to iterate on for other providers.

[DenisaCG]

Merging this now, but we can open a follow-up PR if there is any other aspect that needs to be addressed. Thank you @ellisonbg!