Add demonstration for robust training and robustness quantification

[YuezhuXu]

Mainly three files: examples/tutorials/part4_Lipschitz_estimator for showing how to use package eclipse-nn;
example/control/stabilize_linear_system_with_verification for adding robustness part to Part_1_stabilize_linear_system;
example/control/ref_tracking_ODE_with_verification for adding robustness part to Part_3_ref_tracking_ODE.

[drgona]

These are nice contributions. But we need some modificaitons before they can be merged.

Jupyter notebooks need more markdown to explain the presented methods.
This notebook does a good job at that, but the rest do not.

I see the training is very slow in the reference tracking ODE case with Jabocian regularizaiton, is it the same for the stabilizaiton task?

Let's don't upload any files .mat, .npz, .pt, and others into neuromancer. For all files we should use stable url download.

[drgona]

Ideally, we don't want to store any data files for our examples. If we need to use data from a file we want to load it from stable url. See example in this notebook: https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/DAEs/tank_dae_example.ipynb

[drgona]

The tutorial directory is meant to be only for base neuromancer syntax.
I suggest creating new directory called verification for the Lipschitz estimator files.
The jupyter notebooks should be didactical, the current version has minimal markdown. Please update the new version with each cell explained in markdown.

[drgona]

the function def make_jac_cl_node should be made way more modular. You have multiple function definitions within this function, this could be easily broken apart.