fix: resolve crash when on_csv_write='update' encounters empty column (#1367)

[bhoomikakk]

Fixes #1367

Problem

When on_csv_write="update" is used, pd.read_csv() infers columns with all missing/empty values as float64. On subsequent flushes, attempting df[col].dtype.type(val) fails with ValueError: could not convert string to float when val is a string (e.g. gpu_model="NVIDIA RTX...").

Solution

• Removed df[col].dtype.type(val) and updated DataFrame columns directly.
• Added graceful fallback to object dtype (df[col] = df[col].astype(object)) when assigning values to columns whose pandas dtype cannot hold the value (e.g. string values in an all-NaN float64 column).
• Added unit test test_file_output_out_update_empty_column_second_write in tests/output_methods/test_file.py.

Verification

• Ran unit tests in tests/output_methods/test_file.py (24 passed).