Using pexpect to query wccg. Relies on custom OpenCCG version until shoeffner/wccg-prompt is merged into master.

Adding legacy grammar files until openccg-gum-cooking becomes stable enough to use a more recent version.

Enabling proper flask hot loading by performing a linked installation and running the debug docker config as a script rather than a module.

This resolves #31.

Using port 80 for docker-compose service, reordering Dockerfile for faster rebuilds at the cost of an additionaly RUN.

The development server uses docker-compose run and mounts the
app directory. It then runs the flask development server.
The production server uses docker-compose up and runs flask as a
uwsgi app behind nginx. It does not mount app.