- The dataset API now interprets `float` as 32bit. `double` can be passed as a string to use a floatingpoint with 64bit as dtype (specifying the dtype explicitly, e.g. `np.float64`, still works). The `datasource-properties.json` stores 32bit floats as `"float"` and 64bit floats as `"double"`. [316](https://github.com/scalableminds/webknossos-cuber/pull/316)
- The file `py.typed` was added to tell `mypy` to check this package if it is used in other projects