feat(chunk): export provenance-preserving records

[MMoney1988]

Summary

• Add a chunk-record export path from normalized block JSON to JSONL.
• Preserve source references through block IDs, page numbers, bounding-box refs, source identifier, and source PDF hash.
• Add CLI support via python module and the pdf-quality-chunk entry point.

Why

• Provides a small, deterministic export artifact for downstream review or retrieval-oriented preparation.
• Keeps the scope limited to normalized parser output and provenance preservation.

What

• Added pdf_quality_report.chunk with load/build/write domain functions and a thin CLI adapter.
• Added JSONL chunk records with schema_version chunk_record_v0.
• Added README and walkthrough documentation for the chunk export.
• Added focused tests for grouping, splitting, filtering, provenance preservation, CLI behavior, and example smoke exports.
• Does not add retrieval, embeddings, vector stores, LLM use, UI, OCR, or table reconstruction.

Tests

• .venv/bin/python -m pytest -p no:cacheprovider tests
• .venv/bin/python -m ruff check src tests
• git diff --check
• .venv/bin/python -m pdf_quality_report.chunk --help
• .venv/bin/python -m pdf_quality_report.chunk --input examples/mdpi_pdf_elements/page_006/normalized_blocks.json --output /tmp/page_006_chunks.jsonl
• .venv/bin/pdf-quality-chunk --help

Risks

• The output format is v0 and intentionally small; future format changes should use schema_version.
• Source metadata can only be preserved when present in the normalized input.
• The export does not validate semantic correctness or downstream retrieval quality.

Rollback

• Revert this PR to remove the chunk export module, CLI entry point, tests, and docs.