[codex] sharpen AutoBCI public positioning

[infoechoes]

Summary

• Reposition AutoBCI Harness as a 24/7 research operator for coding agents rather than a BCI product or generic autoresearch clone.
• Add explicit public disambiguation from the UFPA/Vitor Vilas-Boas AutoBCI MI-BCI project and Andrej Karpathy's autoresearch project.
• Add llms.txt and docs/research_forest.md so humans, coding agents, and AI-search systems can extract the intended entity boundary and Research Forest isolation model.
• Update README, AGENTS, the local skill, demo quickstart, and package metadata to keep the public entry points consistent.

Why

Current AI/search behavior can map AutoBCI to an existing UFPA MI-BCI project and AutoResearch to Karpathy's project. This change keeps the repo name but makes the intended entity clearer: AutoBCI Harness is a bounded research-loop operator/control plane with pluggable workers, model providers, fixed evaluators, ledgers, dashboards, and mobile status gateways.

Validation

• git diff --check
• git diff --cached --check
• python - <<'PY' ... tomllib.load(...) ... PY
• PYTHONPATH=src pytest -q tests/test_headless_cli.py tests/test_goal_perp_research_tree.py => 6 passed

Notes

This PR does not change runtime behavior. It intentionally avoids claims that llms.txt or SEO/GEO metadata guarantee AI-search citation; those files only make the public entity boundary easier to extract and test.