COCO superset for proper SemVer

[jaspersiebring]

Feature branch that adds automates SemVer changes to a COCO dataset based on input data (meaning users are no longer prompted to provide versions themselves). This is possible because we now keep track of the raster sources used to make the image subsets that contained annotations, technically making it a COCO superset (i.e. GeoCOCO datasets have all COCO-required keys plus others). It is still compatible with any COCO-accepting data applications though.

New COCO datasets are tagged as '1.0.0`, indicating that all initial annotations are made from the same raster source and stored in the same output directory.

• PATCH: If this same dataset is used again to store (updated) annotations from the same raster source in the same output directory, we treat it as a patch and bump the SemVer version by 0.0.1.
• MINOR: If this same dataset is used again to store annotations from a new raster source in the same output directory, we treat it as a minor release and bump the SemVer version by 0.1.0.
• MAJOR: Lastly, if this same dataset is used again to store annotations from any raster source (new or existing) in a new output directory, we treat it as a major release and bump the SemVer version by 1.0.0. That is because it technically introduces breaking changes as you now have n+1 output directories to keep track of.

It is also the first branch released under the Github Actions based CI meaning that all (existing and new) code is tested and checked for formatting and sufficient test coverage (using mypy, pytest, pytest-cov, ruff).