Bug: Resource Leak in CSVHandler (Unclosed File Handles)

[debanshd]

In csv_handler.py, the _read_streaming and _header methods use PyArrow's pa_csv.open_csv by passing a string path. When count_rows=False (the default!), the file stream is read for schema retrieval and then abandoned without being exhausted. This can lead to file descriptor exhaustion in batch processing environments where many CSVs are analyzed sequentially without exiting the process.

Affected Location:

src/croissant_maker/handlers/csv_handler.py

    @staticmethod
    def _read_streaming(file_path: Path, convert_options, count_rows: bool = False):
        try:
            reader = pa_csv.open_csv(str(file_path), convert_options=convert_options)
        except UnicodeDecodeError as exc:
            # ...
        
        # If count_rows is False, we return without exhausting the reader!
        if count_rows:
            # exhausted
        else:
            num_rows = None # Leaves file open in memory