Introduces the core structure and functionality for the Earth Observation Machine Learning (EOML) project.

Key additions:
- **Automation**: Implemented task management, and TOML-based configuration for experiments (`automation/`).
- **Data Handling**: Added support for geospatial data structures, LMDB persistence, and serialization (`data/`).
- **Raster Processing**: implemented efficient raster reading and window-based sample extraction (`raster/`).
- **Deep Learning**:
  - Added PyTorch model factories and custom CNN architectures (`torch/models.py`, `torch/resnet.py`).
  - Implemented dataset loaders for training and mapping (`torch/cnn/`).
  - Added training loops and validation utilities (`torch/trainer.py`).
- **Scripts**: Included entry points for land cover mapping and mosaicking (`bin/`).
- **Infrastructure**: Added project configuration (`pyproject.toml`, `.gitignore`) and IDEA settings.