Instead of Classifier.

Also update the codebase accordingly.

Some design rework needs to be done first.

- Parameter min and max of min_max_scale are now the intended final range
- Add clip parameter to min_max_scale stdlib function (if enabled, clipping is
  also done after the added perturbation)

- Add support for neural network as array in language.ml
- Translate csv record as vnnlib constraints in dataset.ml
- Add simple dataset normalization functionalities in dataset.ml
- Add vnnlib printer (as smtlib printer with direct real support)
- Add vnnlib driver for PyRAT
- Rework onnx and nnet parsers to not fail on non-mandatory information

Include building a proper Why3 float literal.

Also, consider SAVer as complete wrt "correct" predicate, which does not take
into account a perturbation anymore. Correctness is wrt the expected label only.