docs: add detailed README with technical plan for tarot teleprompter web app

[ID377585]

Motivation

• Provide a concise, actionable technical plan to build a web-based teleprompter that recognizes tarot cards via camera and displays private interpretations on a secondary device.
• Capture architectural decisions that prioritize a reliable live experience (offline-first PWA, local inference, simple UX) and a clear MVP path.
• Specify practical recognition and data strategies so the project can progress from prototype to robust product without speculative choices.

Description

• Replaced the placeholder README.md with a comprehensive, Portuguese-language guide that defines goals, two-device flow, and the ideal user experience for a teleprompter-style tarot reader.
• Documented a recommended stack and APIs: TypeScript + React + Vite, getUserMedia/enumerateDevices, Fullscreen & Wake Lock APIs, TensorFlow.js (Teachable Machine) for in-browser classification, PWA + Workbox, and IndexedDB for local storage.
• Added practical recognition strategy (MVP: TFJS classification, 224×224 frames, stabilization voting, handling inverted cards), content schemas (cards.json, spreads.json), and a Registrar Cartas capture workflow for collecting training images.
• Included minimal MVP feature list, phased roadmap, suggested project folder structure, and product decisions focused on reliability in live streams.

Testing

• Validated the change by viewing diffs with git diff -- README.md | sed -n '1,220p' which succeeded.
• Confirmed working tree status with git status --short which returned expected modified file.
• Printed and inspected the final README.md content via nl -ba README.md | sed -n '1,260p' to verify the inserted documentation.

---

Codex Task