Add GitHub Pages docs site + AI-search discovery layer (Phase 1)

[UnscriptedByKraMy]

Build a MkDocs (Material) documentation site from the repo's own markdown so AI crawlers and search engines can index the project, and publish the llms.txt-family files for LLM consumers:

• README SEO quick wins: intro names both "AI Engineering Lab" and "Provider-Aware Statusline", hosted-site pointer, FAQ, repo-map row, absolute tests-badge URL
• mkdocs.yml: Material dark theme, explicit nav, strict build
• requirements.txt: pinned mkdocs + mkdocs-material
• robots.txt: allow all crawlers, point at sitemap
• Scripts/assemble-site-source.js: stage tracked docs + assets into a gitignored docs/ (MkDocs requires docs_dir to be a child directory)
• Scripts/generate-llms.js: generate llms.txt + llms-full.txt from the same source markdown, links rewritten to absolute site URLs
• CI: new docs job runs mkdocs build --strict on every push/PR
• Deploy: deploy-site.yml publishes the site to GitHub Pages on main

The repo root stays the single source of truth; docs/ and site/ are gitignored staging/build output.

What does this change?

Brief description.

Evidence

This project values evidence over assumption. For behavior changes, describe what was
observed or which authoritative source supports the change.

Checklist

• npm test passes (all 68 contract tests)
• No machine-specific identifiers added (absolute home paths, usernames,
  hostnames, local gateway ports)
• node Scripts/check-markdown-links.js passes (if markdown changed)
• New source files carry the // SPDX-License-Identifier: MIT header
• No secrets, tokens, prompts, or request content included