Functionality for machine learning data generation

This release provides functionality for generating data for training machine learning models by being able to output data for the full domain over time for multiple source positions in an easy and compact way. The setup file includes extra fields for handling the data creation, such as various ways of defining source positions (read from vertices in a .msh file or explicit given in setup file), source types (Gaussian pulse or Gaussian random fields), output data temporal resolution, and more. Also, impulse responses at specific locations can be generated as in the versions of the previous code, but now .wav files are generated with temporal resolution denoted in the settings file.

All data used for training the models from the paper Sound propagation in realistic interactive 3D scenes with parameterized sources using deep neural operators by Borrel-Jensen et al. can be reproduced with this version.