right now I think we're assuming every notebook we run will be on S3, but yet we allow saving the output notebooks to local disk. Let's allow running local notebooks as well.
I am thinking the /run endpoint should get a new parameter called location that can be either s3 or local for now. This also allow us to add other providers in the future.