feat: Share PDF page extraction across PDF ingest paths

[gwokhou]

Summary

Refs #135.

This PR introduces a shared PDF page extraction contract and routes PDF ingest paths through it where doing so preserves existing behavior:

• adds PageContent, PdfExtractor, LocalPdfExtractor, pages_to_markdown, and pages_to_json
• routes short PDF Markdown conversion through the configured shared extractor
• routes long PDF local fallback page JSON through the same extractor
• preserves PageIndex Cloud get_page_content() as the first choice when PAGEINDEX_API_KEY is set, then falls back to the configured extractor
• updates tests around PDF extraction, short PDF conversion, long PDF source JSON, and mocked async compile coroutines

Reviewer Note

Issue #135 asks for a configured PDF extraction engine to affect both short PDF Markdown and long PDF page JSON. This PR preserves the existing PageIndex Cloud page-content path when PAGEINDEX_API_KEY is set because that path provides cloud OCR behavior for scanned or complex PDFs.

That means pdf_parser controls short PDFs and long-PDF local fallback JSON, but does not override cloud-mode get_page_content() output. Please confirm whether preserving cloud OCR priority is the desired compatibility behavior, or whether #135 should require pdf_parser to control cloud-mode long PDF page JSON too.

Validation

UV_CACHE_DIR=/tmp/uv-cache UV_PYTHON=3.13 uv run --extra dev pytest tests/test_pdf_extractor.py tests/test_converter.py tests/test_indexer.py tests/test_config.py tests/test_add_command.py tests/test_remove.py tests/test_url_ingest.py

Result: 161 passed, 6 warnings in 1.35s.