Changes visualize driver functions to return graphviz object

This change is backwards compatible. It returns and wires
through returning the graphviz object created. That way in a notebook,
it can render the DAG for you.

It also makes saving to a file optional -- you just need to
pass in None explicitly. Adds test for it.

The hello world example is also updated with a real notebook
and shows Thierry Jean's approach of using the autoreload
extensions.

examples/hello_world/README.md

@@ -9,9 +9,9 @@ File organization:
 * `my_functions.py` houses the logic that we want to compute. Note how the functions are named, and what input
 parameters they require. That is how we create a DAG modeling the dataflow we want to happen.
 * `my_script.py` houses how to get Hamilton to create the DAG and exercise it with some inputs.
-* `my_notebook_script.py` houses how one might iterate in a notebook environment and provide a way to inline define Hamilton
+* `my_notebook.ipynb` houses how one might iterate in a notebook environment and provide a way to inline define Hamilton
 functions and add them to the DAG constructed. To be clear, it is not used by `my_script.py`, but showing an alternate path
-to running things.
+to running/developing things.
 
 To run things:
 ```bash