Add h5 loader

[Haydnspass]

Do it similarly like this, such that we can load decode emittersets.
Convert to nm if xyz was in px units.

def load_h5(path) -> Tuple[dict, dict, dict]:
    """
    Loads a hdf5 file and returns data, metadata and decode meta.

    Returns:
        (dict, dict, dict): Tuple of dicts containing

            - **emitter_data** (*dict*): core emitter data
            - **emitter_meta** (*dict*): emitter meta information
            - **decode_meta** (*dict*): decode meta information

    """
    with h5py.File(path, "r") as h5:
        data = {
            k: torch.from_numpy(v[:])
            for k, v in h5["data"].items()
            if v.shape is not None
        }
        data.update(
            {  # add the None ones
                k: None for k, v in h5["data"].items() if v.shape is None
            }
        )

        meta_data = dict(h5["meta"].attrs)
        meta_decode = dict(h5["decode"].attrs)