fix: preserve converted PDF URL when hydrating parse results

[Kayvan-Zahiri]

When handle_url_response() hydrates a URL result, it reconstructs FullParseResponse without copying pdf_url. The same helper preserves that field for inline results. A caller therefore loses the converted PDF link depending on response size.

Copy response.pdf_url into the hydrated response. Four offline regressions compare inline and URL results with a signed PDF URL and with no URL, also checking that hydration retains the document text and other existing helper metadata.

Validation:

• Before: 1 regression fails and 3 controls pass; afterward all 4 pass.
• Pydantic 2 offline suite: 457 passed, 2 skipped.
• Separate Pydantic 1 offline suite: 450 passed, 9 skipped. The helper already uses Pydantic 2 APIs; the 4 new helper cases are explicitly skipped under Pydantic 1.
• Excluded tests/api_resources (requires the API mock server) and the default-excluded live tests/e2e. No live API or storage requests.
• Repository Ruff lint, changed-file formatting, mypy, pyright, and git diff --check pass.
• Broader mypy and source/test pyright retain dependency/runtime typing failures reproduced with the original helper (4 and 22 respectively). No generated runtime changes.

---

^{Need help on this PR? Tag @codesmith-bot with what you need.}