implement a surrogate model example using NDK data

[d-lucena]

Experiment with machine learning models to evaluate their accuracy in reproducing simulation results given the transducer and material properties.
This would be an initial implementation with a simplified scenario as a POC:

• only one supported transducer
• only a section of the brain is simulated (to reduce computational time for training data collection)
• keep a (near-)fixed distance from transducer to skull
• only 2D simulations
• only using the skull already available in NDK