* Modified `evaluate_network` to process nets with multiple outputs, but still only returns 1 output (first output of given NN)
* Added 2D relu function (left 1D relu for HornNets)
* Added function to handle classification NNs
  * `discretize_classifier_output` converts real number output to "indices" based on input thresholds
  * `impose_classifier_constraint` multiples classifier output column with other columns based on map
* Extended character length of network filename variable to allow longer names
* Added names, default settings, load function, evaluate function, CLI, and make instructions for new network named `adept`
* Added git-LFS pointers to new network namelist files
* Updated `data/README.md` to include new repository for `adept` weights and biases (currently a private repo)

Merge branch 'aaronkho/active_learning_nn' into 'main'
See merge request qualikiz-group/QLKNN-fortran!48

Name is preliminary due to lack of creativity but subject to future change

The output array of the function is still 1D, returning only the first output of NN

Left 1D relu function as it is needed for HornNets

* Removes the testing suite at EPFL. Integration tests will stay in RAPTOR
* Adds a working MATLAB/MEX testing pipeline to the PSNC gitlab
* Adds deeper Fortran tests for MEXed parts of QLKNN
* Adds buggy(!) implementation of QLKNN-HornNet to mexed interface; to be tested! Known bugs in there right now!
* Adds a Jacobian checker in pure Fortran; based on the RAPTOR Jacobian tests
* Adds tools to write namelists in MATLAB

Merge branch 'karel/mex-hornnet-rebase-rebase-eufus' into 'main'
See merge request qualikiz-group/QLKNN-fortran!46

This reverts commit eea6789b.
To be re-enabled later