Implement %callflow magic to spawn CallFlow app directly on ipython notebooks.

[jarusified]

The interface possibly is going to mimic Tensorboard's style of launching a web-interface from a notebook interface.
See tensorboard

import callflow

%load_ext callflow

%callflow --config /path/to/config/file

• Mimic the style used in Tensorboard.
• Spawn a python and npm process to run on different ports.
• Send data through the interface and launch the view.
• Ensure no-duplication of processes.
• Add documentation.
• Adjust the aspect ratio of the tool view, if possible. (This will be part of different ticket).

Progress:

• Current implementation fails to launch a subprocess for python3 server/main.py, complains no such file. I believe we need to set some environment variables to launch the required processes from the correct directory.

Useful Links:

• Allow Jupyter+TensorBoard to work w/ hostname other than localhost tensorflow/tensorboard#2407
• Support jupyter-server-proxy (and other proxies) tensorflow/tensorboard#3674
• https://github.com/jupyterhub/jupyter-server-proxy