Add SKIPP'D Dataset

[nilsleh]

This PR adds the SKIPPD dataset and datamodule with additional information available in their repo. The task is to predict power output from fish-eye images of the sky. They have done two approaches from my understanding:

1. "Now Cast": Predict the power output from a single RGB image
2. "Short-term Forecast": from time-series images make a forecast

This dataset version implements approach 1. Once we support Forecasting in torchgeo, I would add the "forecast" version of this dataset.

Dataset Format:

• .hdf5 file containing images and labels
• .npy files with corresponding datetime timestamps

Dataset Features:

• fish-eye RGB images (64x64px)
• power output measurements from 30-kW rooftop PV array
• 1-min interval across 3 years (2017-2019)

The dataset comes with a "trainval" and "test" split:

• For the datamodule I implemented a 80/20 random split on the "trainval" split to generate a train and validation set

Examples: