Updating of examples and docs

The examples used to train RF should be updated. As really the main one we want users to see is the `nd_to_md_regression_direct_withcov.jl` This will provide additional clarity of how the tools work.
- [ ] Ensure all examples work with the latest RF version (currently this is v0.3.0)
- [ ] Remove experimental examples for hyperparameter training, leaving those approaches that are promising (e.g. `nd_to_md_regression_direct_withcov.jl`)
- [ ] Create documentation page explaining the hyperparameter training example. With a walkthrough of why HP training is needed and how it is approached here, and the fact that it works on many examples.