Sync evaluate_from_hf.py: filter redundant/decomposable boxes and stable-sort ground truth to match loading.load_annotation_boxes

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

Add requirements.txt, fix split default

Delete evaluation_kit/box_grouping/__pycache__/loading.cpython-312.pyc

Delete evaluation_kit/box_grouping/__pycache__/spatial.cpython-312.pyc

Delete evaluation_kit/box_grouping/__pycache__/models.cpython-312.pyc

Delete evaluation_kit/box_grouping/__pycache__/group.cpython-312.pyc

Delete evaluation_kit/box_grouping/__pycache__/geometry.cpython-312.pyc

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

Delete evaluation_kit/__pycache__/evaluate_from_hf.cpython-312.pyc

Delete evaluation_kit/__pycache__/convert_predictions_to_evaluation_csv.cpython-312.pyc

Add standalone evaluation kit