Examples for HyperParameter Tuning in Learning to Rank tasks with Python API

I'm trying to run MMLSpark with LightGBM algorithms and applying hyper parameters tuning, but it is not so easy to adapt it based on the actual examples and with the Python API.

I would like to see some examples where the learning to rank algorithms are successfully applied and the parameters are also tuned with TuneHyperparameters class.

Additionally: 1) the different options of input datasets would be presented, e.g. libsvm, Parquet, etc; 2) integration with other optimization libraries would be a plus, e.g. [HyperOpt](https://github.com/hyperopt/hyperopt), [RAPIDS](https://rapids.ai/start.html), [Optuna](https://github.com/optuna/optuna/blob/master/examples/lightgbm_simple.py), ...

Start with the TuneHyperparameters class and the Python API usage for LightGBM learning-to-rank tasks. Review how libsvm and Parquet datasets are represented, then define examples that demonstrate successful ranking and parameter tuning; done means the requested examples are usable and cover the stated dataset options.

Written by the indexing model from the issue text.