Tolerate datasets with fewer optional annotation fields; add guillemet normalization

Extract shared line-grouping/reading-order helper in prediction.py

Delete evaluation_kit/evaluation/__pycache__/text_metrics.cpython-312.pyc

Delete evaluation_kit/evaluation/__pycache__/reports.cpython-312.pyc

Delete evaluation_kit/evaluation/__pycache__/prediction.cpython-312.pyc

Delete evaluation_kit/evaluation/__pycache__/measure_overall_accuracy.cpython-312.pyc

Delete evaluation_kit/evaluation/__pycache__/measure_accuracy.cpython-312.pyc

Delete evaluation_kit/evaluation/__pycache__/generate_accuracy_report_variants.cpython-312.pyc

Add standalone evaluation kit