Rebuild voSINT v2: token-free Playwright adapters, modular CLI, Pinterest-first default

[Meshall]

Motivation

• Replace the old SerpApi/config.ini/token workflow with a simpler, token-free architecture that is friendlier for normal users and stronger for OSINT provenance workflows.
• Make Pinterest the default provider and use browser automation to support robust, multi-provider scraping without requiring API keys.
• Provide a structured, testable codebase and a simple CLI so workflows (scan/batch/diff/report) are repeatable and auditable.

Description

• Reorganized the project into a vosint package with core, providers, models, output, and utils modules and added pyproject.toml and .env.example.
• Removed SerpApi/config.ini and legacy modules/* logic and implemented Playwright-based provider adapters (Pinterest default, Google Lens, Bing Visual, Yandex, TinEye) behind a provider registry and ProviderAdapter API.
• Implemented multi-frame extraction and ranking, OCR/transcript pivot hooks, VideoDNA, Hit/Case/FramePivot models, normalization/dedup, timeline ranking, confidence scoring, and case-folder outputs (HTML/JSON/CSV + raw/normalized artifacts).
• Added an argparse CLI (vosint scan|diff|report) with fast|deep|stealth modes and flags for providers, OCR, transcribe, frame limits, browser/headful controls, outputs, and keep-frames behavior; also added unit tests and README updates.

Testing

• Ran the unit test suite with python -m pytest -q, which passed (2 tests, all green).
• Verified the CLI help with python voSINT.py --help, which executed successfully and displayed the new commands and flags.
• Performed a smoke run of vosint scan <video> --mode stealth --max-frames 1 --no-browser which executed the orchestrator path and produced case artifacts (frame extraction, report files); a separate Playwright screenshot attempt failed due to the container/browser environment (SIGSEGV), which is an environment limitation and not a functional regression in the code changes.

---

Codex Task