Although the majority of common analysis use cases requires only the parsing of Harp flat binary files and messages into data frames, it is also desirable to support the converse operation: generating Harp binary files from existing data frames.

This PR enables post-processing and storage of processed datasets using the Harp binary format, for compatibility with existing systems. Both a write and format functions are defined to allow writing data to a Harp binary file, or to a memory array buffer, respectively. Regression tests were added to validate successful round-trip. Example signature for the write function:

def write(
    file: Union[str, bytes, PathLike[Any], BinaryIO],
    data: pd.DataFrame,
    address: int,
    dtype: Optional[np.dtype] = None,
    port: Optional[int] = None,
    epoch: Optional[datetime] = None,
    message_type: Optional[MessageType] = None,
):

Further testing may be required to ensure timestamp roundtrip is not lossy if done over float64. Given the specs of the 8-bit protocol this should be the case, but conversion may require revision in case of the 32-bit protocol.