Proposal for parse_tsv_matrix refactoring

[laurent-laporte-pro]

The parse_tsv_matrix could be efficiently refactored like that:

def parse_tsv_matrix(file_data: bytes) -> List[List[MatrixData]]:
    """
    Parse a TSV file data and return a matrix of data.
    
    Args:
        file_data : The TSV file data
    
    Returns:
        The matrix of data.
    """
    file = io.BytesIO(file_data)
    reader = csv.reader(file, delimiter="\t")
    return [[MatrixData(elm) for elm in row] for row in filter(None, reader)]

[sylvlecl]

Actually, using pandas to read those CSVs will be a LOT more efficient than a for loop.
It means changing the return type, though, so it has some impact. A plain List[List[MatrixData]] is a very poor representation, in python, for a numeric matrix, especially for the typical 8760x200 matrices that we use!

It's already done this way in raw study services, it seems: not sure why it's not homogeneous ?
See here:

AntaREST/antarest/study/storage/rawstudy/model/filesystem/matrix/input_series_matrix.py

Lines 40 to 45 in d6eae53

	def parse(
	self,
	file_path: Optional[Path] = None,
	tmp_dir: Any = None,
	return_dataframe: bool = False,
	) -> Union[JSON, pd.DataFrame]: