Ep-bolfi

This major update to EP-BOLFI contains automization-suitable scripts for the execution of the optimization and utility routines that were already there. These are written to be used with the database software [Kadi4Mat](https://kadi.iam.kit.edu/), but you may adapt them to other storage solutions with minimal changes. Additionally, some data operations may be automated such that no script is needed for an entire parameterization of timeseries or impedance data.
- GITT parameterization workflow, based on the comparison of square-root behaviours between measurements and PyBaMM models.
- Draft of an impedance parameterization workflow, based on Distribution of Relaxation Times features and an analytic solution of the SPMe (and hence, DFN) model.
- Interactive exploration of model parameter influences on impedance.
- Extraction of measurement protocol, dynamic behaviour features, and static OCV information from timeseries measurements.
- Extraction of overpotential from timeseries voltage data.
- Apache Parquet file format support for compact handling of measurement data.
- Various utility functions for dataset transformations, spline fitting, and visualization.

For the BASF battery dataset used in the examples, please find it at https://zenodo.org/records/7478267.

Install with `pip install ep-bolfi`.

Fixed several issues that prevented the example code from running:
- The missing file "examples/parameters/models/ocv_curves.py" was added.
- All instances where data files were read in now point to the correct path.
- Any instance of the √ character in console output was replaced by "sqrt".

A slight code error prevented EP-BOLFI 2.0 and 2.1 from working with multiple estimation parameters. With this hotfix, that functionality is restored.

To be able to just `pip install ep-bolfi`, I updated the install files accordingly. This is the first version that appears on PyPI.

This release contains EP-BOLFI in Open Source, ready to be incorporated into other libraries.

There are also some improvements over the version that was released Closed Source initially:
- I added better heuristics to EP, which make EP-BOLFI run more reliable and allow for more fine-tuning.
- Battery model handling now enables the simulation of three-electrode setups.
- The OCV fitting procedure, an implementation of the Birkl2015/MSMR model, now produces even better and more consistent fits.
- Data handling procedures now provide export to JSON and Apache Parquet.
- The examples do no longer depend on self-implemented battery models.

Complete code base and data necessary to reproduce the results in the associated paper.
May be used with custom simulators and GITT data to parameterize said GITT data.