Add standalone evaluation (no Arena CLI needed) + update program.md

- scripts/standalone_eval.py: Run OfficeQA questions using Docker directly
  - Uses public corpus image (ghcr.io/sentient-agi/harbor/officeqa-corpus)
  - Runs OpenHands SDK agent inside container
  - Scores with 1% fuzzy tolerance
  - Resume support (saves after each question)
  - Filter by UID, limit count, or run all 246
- program.md: Updated for standalone flow (no Arena CLI dependency)
- .env.example: Simplified for standalone use

Friend can now: clone → docker pull corpus → set API key → run eval

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>