feat(app): implement end-to-end vanilla agent workflow

[talhasarit]

This pull request implements the complete, end-to-end vanilla agent workflow as defined in the project's technical specification. It
brings the application from a scaffold to a functional state, allowing a user to input a project idea and receive a generated PRD
streamed to the UI in real-time.

This resolves the core requirements for Phase A of the project roadmap.

Key Changes:

• Backend:
  • Implemented the VanillaAdapter to make live calls to OpenAI and Google LLMs.
  • Implemented the full agentic pipeline (outline → draft → critique → revise) with detailed, task-specific prompts.
  • Replaced the InMemoryStore with a persistent RedisStore for robust state management.
• Frontend:
  • Connected the Streamlit UI to the backend.
  • The UI now initiates a generation run via a POST request and opens an SSE connection to stream and display live updates (status,
    content, and diffs).
• Documentation:
  • Updated the README.md to accurately reflect the project's current implementation status.
• Code Quality:
  • Resolved all pre-commit hook failures, including all mypy type-checking errors, to ensure a clean CI run.