[lenzhq]

Summary

Adds client.extractFile(file, options) to the Node SDK for POST /api/v1/extract/file (lenzhq/Lenz#1585), keeping the SDK version at 3.6.0 (3.5.0 is the latest published version on npm).

• client.extractFile(file, options) accepts Blob | ArrayBuffer | Uint8Array for .docx, .pdf, .png, .jpg, .jpeg, and .webp files (up to 10 MB).
• Supports filename, language (defaults to "auto", the document's own language), focus (up to 300 characters), locate (returns passage text in positions[i].text with start/end: null), idempotency / idempotencyKey, and per-call request options (timeoutMs, maxRetries, signal, headers).
• Shares the 150s default timeout floor (EXTRACT_TIMEOUT_MS) with extract(), and preserves /extract/file review-family error codes in src/compat.ts.
• Exports ExtractFileOptions from both src/index.ts and src/index.browser.ts.
• Refreshes openapi.json with /api/v1/extract/file and updates README.md and CHANGELOG.md ([3.6.0]).

Reviews

• Astra review (codex review --base origin/main): Reviewed e97be28; flagged openapi.json emitted without API_CANONICAL_ACCOUNTS_FROM (fixed in 69b6821).
• Opus review (claude -p --model opus): Reviewed e97be28; flagged isReviewFamily missing /extract/file in src/compat.ts, client-side 20 MB cap in resolveUpload, language docstring default, and test assertions (all fixed in 69b6821).

Post-deploy actions and prod testing

1. Deploy lenzhq/Lenz#1585 first so POST /api/v1/extract/file is live in production.
2. Merge this PR (version stays 3.6.0 until ready to tag v3.6.0 for npm publish).
3. Smoke-test await client.extractFile(pdfBytes, { filename: "sample.pdf", locate: true }) against https://lenz.io.