Gaitmap-datasets

Fixed

- The convert event methods for the EgaitAddidas2014 dataset had a bug that the offset was not taken into account
correctly.
- For two participants of the EgaitAddidas2014 dataset the offset between the systems was for some reason not consistent
across all strides. We decided to ignore this, but warn the user about it.

Changed

- The `mocap_sampling_rate_hz` method of the Kluge2017 dataset is now called `mocap_sampling_rate_hz_`
(with trailing underscore) to be consistent with the other datasets.

Added

- SensorPositionComparison2019 now has a method to get the marker trajectories per stride.

Fixed
- The egait segmentation validation dataset contained strides that had negative duration.
We now remove them during loading.

Added
- New labels for the EgaitSegmentationValidation dataset have been created and are distributed with all future versions
of the dataset.
These new labels are now the default labels used by the dataset.
The old labels can be accessed using the `segmented_stride_list_original_` property.

Fixed
- Fixed further potential cases where the index of datasets was not deterministically sorted.


Changed
- Config files are[CHANGELOG.md](CHANGELOG.md) frozen now

Fixed
- Fixed an issue that the index of `EgaitParameterValidation2013` was not deterministically sorted.

Changed

- Removed pydantic as dependency and switched to custom logic to set configs.
The use-facing API remains the same.

Added

- The `EgaitParameterValidation2013` dataset now has the option to exclude participant "P52", which appears to have
non-sensical imu data.
- The `EgaitAdidas2014` dataset was added.
The dataset is not publicly available yet, but will be soon and contains data from healthy participants walking
through a Vicon motion capture system.
- A `convert_segmented_stride_list` was added to convert segmented event list to min_vel eventlists.

Changed

- The method `convert_with_padding` of the `SensorPositionComparison2019` dataset is now called
`convert_events_with_padding`.