docs: add benchmark evidence to org profile; register openadapt-bootstrap

[abrichr]

What changed

• Adds an Evidence section to profile/README.md, between the quickstart and the six-surface map: the live OpenEMR field result and the CI-reproducible MockMed control, each stated with its published caveats and a pointer to LIMITS.md.
• Registers openadapt-bootstrap as Internal in repository-lifecycle.yml and REPOSITORY_LIFECYCLE.md; it was previously unclassified.

Why

The merged launch-truth profile (#2) gives visitors the thesis, quickstart, surfaces, and lifecycle labels — but no evidence. The website leads with the benchmark headline; the org front door should carry the same story with the same honesty. This strengthens #2's direction without changing structure, markers, or required links.

Sources for every claim

• 20/20, 39.2s vs 70.4s p50, $0 vs $0.5522 model cost/run: benchmark/openemr/BENCHMARK.md at OpenAdaptAI/openadapt-flow@main (results table)
• 100/100 vs 20/20, 4.9s vs 37.5s p50: benchmark/BENCHMARK.md at OpenAdaptAI/openadapt-flow@main (results table)
• "small sample / shared daily-resetting demo / not CI-reproducible", "cost and latency, not success rate", and the extrapolation caveats: "Reading the benchmark evidence" in docs/LIMITS.md at OpenAdaptAI/openadapt-flow@main
• Internal label for bootstrap: repository description ("Self-hosting infrastructure for OpenAdapt recursive development") — team tooling, not a public product surface, per the registry's own definition of Internal

Validation

• python3 scripts/check_profile.py → "Validated canonical product truth and 27 Markdown links." (canonical truth string, required markers, and required links all preserved)

🤖 Generated with Claude Code