Mlpf

* add custom progress bar for examples
* remove ascii=True flag from every tqdm progress bar
* add Min and Max metrics

* Refactor power flow errors in such a way that the calculation is now
separate for active and reactive power decoupling them
* Refactor TorchMetrics custom metrics to return only one metric

* rename bounds error functions
* move PF/OPF mask functions
* add option to solve OPF when generating data
* extract OPF data
* create metrics for PF and OPF
* unite numpy/torch data objects
* refactor ppc -> (o)pf conversion by splitting it into multiple smaller functions
* removed the functions that create the data objects. Instead, class constructors do that
* remove opf_feature_matrix from OPFData and replace it with the corresponding fields themselves

* Configure setup.py such that `pip install mlpf` works.
* Separate torch and numpy in data conversion

Add README.md as long description in PyPi.

First release