DR-057 — PDF form-field detection adopts commonforms directly (licence ratified Apache-2.0)

The owner confirmed the commonforms code licence as Apache-2.0 (2026-07-12), closing the S466 open question: PDF field detection uses the commonforms package directly (real procurement PDFs are FLAT — 0 AcroForm — so detection is required), paired with pdfplumber label pairing (detection yields coords/widget only). The ARCH-REVIEW §4.3 R-B in-repo ~200-line pypdf+onnxruntime wrapper over the FFDetr weights is DROPPED as a fallback. Provenance: ID-145 FORM-EXTRACTION-SPIKE.md; owner ratification S467; {145.3} TECH.md.

Substance now lives in: canonical:scripts/cocoindex_pipeline/form_extractors/pdf.py