Training CGL on the GAP loss using the projected data set method

[o2i75nyv942757]

At this point, we have confirmed the romnet is capable of matching results previously collected using the Noack Model. It would be nice to now train romnet on the Complex Ginzburg-Landau equation.

Plan:

1. Similar to training the Noack model, one should create two methods: rom() and test_rom(). Here, rom() generates the reduced order model trajectories and test_rom() will evaluate the normalized l2 error, plot the 2-dimensional output, and plot the real/imaginary full state output. Furthermore, we will use the projected dataset method seen in cgl.py.
2. Train the CGL model on the GAP Loss using the cluster training code developed for training the Noack Model. While romnet trains, finish pitchfork model example in the JNS paper.

After:

1. Contact Sam regarding axisymmetric jet flow data to begin training on this new problem.