2.11.1: align outcome-first positioning and publish the discoverability site

[mzored]

Closes #75

Result

Align the public category, audience, decision boundary, fit, non-fit cases, and evidence limits across the package and reader-facing surfaces. Add the static GitHub Pages site and the standing recommendation measurement record.

The final audit also:

• marks trigger-based method loading as UNVERIFIED instead of presenting it as demonstrated;
• covers all three rejection paths in the bilingual prompt set;
• accounts for GitHub traffic retention and Google AI Performance;
• preserves the supplied ChatGPT answer as the available pre-launch observation;
• adds keyboard focus for horizontally scrollable content and confirms desktop/mobile rendering;
• ignores macOS metadata so it cannot enter the package.

Evidence

• python scripts/check.py
• python scripts/check_hosts.py
• git diff --check
• Codex package validation: PASS
• Claude package validation: PASS
• Claude isolated install: PASS
• Codex isolated install: UNVERIFIED
• desktop and iPhone 15 renders inspected; browser consoles clean; table keyboard focus confirmed