feat: initialize pakdocling core architecture, schemas, preprocessor,…

[milliyin]

Summary

This PR initializes the core architecture for pakdocling (Pakistani Document Intelligence Library) at version 0.0.1, delivering a Docling-aligned public API for structured data extraction from Pakistani identity and educational documents into validated Pydantic JSON objects.

Closes #13
Closes #6

---

🚀 Key Changes

1. Docling-Aligned Public API (pakdocling)

• DocumentConverter: Primary converter class matching Docling's DocumentConverter interface (converter.convert(source, doc_type="cnic")).
• convert(): Top-level functional helper for 1-line conversions (convert("sample.png")).
• ConversionResult: Result object with .document (Pydantic model access), .export_to_json(indent=2), and .export_to_dict() methods.
• Backward Compatibility: DocumentPipeline, extract_document(), and ExtractionResult aliases maintained.

2. OpenCV Image Preprocessing & Modular OCR (pakdocling.preprocessing & pakdocling.ocr)

• OpenCV Pipeline: Grayscale conversion, edge-preserving denoising (cv2.fastNlMeansDenoising(h=10)), CLAHE contrast enhancement (clipLimit=2.0), document deskewing, and adaptive thresholding.
• Bounding Box Rectangles: Added ImagePreprocessor.annotate_boxes() to draw green bounding boxes around detected text regions on original images.
• EasyOCR Integration: Production EasyOCREngine with Urdu & English default languages (['ur', 'en']), lazy loading, and min_confidence=0.2 noise filtering.
• Mock Engine: Fast MockOCREngine for ultra-fast offline unit testing (<10ms) without model downloads.

3. Document Schemas & Models (pakdocling.models)

• CNICData: 13-digit CNIC number, full name, father/husband name, gender (with NADRA odd/even validation), DOB, issue & expiry dates, format variant (old_green vs new_blue Smart Card).
• MatricCertificateData: Roll number, registration number, student & father name, BISE board, passing year, total & obtained marks, grade, group.
• IntermediateCertificateData: HSSC roll number, BISE board, student & father name, total & obtained marks, grade, group (Pre-Engineering, Pre-Medical, ICS, Commerce, Humanities).
• UniversityDegreeData: Degree title, major, university name, student & father name, registration number, graduation year, CGPA.

4. Extractors & CLI (pakdocling.extractors & pakdocling.cli)

• Layout & pattern extractors: CNICExtractor, MatricExtractor, IntermediateExtractor, DegreeExtractor.
• Typer CLI with pakdocling info, pakdocling convert, and pakdocling extract commands.

---

🧪 Verification Results

• Unit Tests: pytest (20/20 passed in 0.28s).
• Linter & Formatter: ruff format --check . and ruff check . passed (0 issues).
• Type Checker: mypy pakdocling (0 type errors across 15 source files).
• CI Matrix: GitHub Actions checks passing on Python 3.10, 3.11, and 3.12.