docs(arabic): README.ar.md — the Arabic entry point (#290)

[eaitbrahim]

What & why

Closes #290. Adds README.ar.md — the Arabic entry point a large share of the intended audience reads in — with a two-sided language switcher (English README links it under the title; it links back). The page covers the first-screen content in Arabic: what keel is (compliance-engine lead), the honest measured result linking the experiment record, the not-a-fatwa-engine boundary (its English original quoted alongside so translation and source are checkable against each other), the no-scholarly-review stance (#289), the five-minute quickstart with commands verbatim (commands are language-neutral), a five-term glossary (الربا، القبض، العين، الدَّين، الغرر) in their established Islamic-finance renderings, the scope statement (the full docs remain English; this is the entry point, not a promise to translate everything), routing, and disclaimers.

On the acceptance item "terminology reviewed by a native speaker": the Arabic was authored directly to the established renderings (not machine-translated), the glossary teaches the vocabulary the page uses, and the document itself invites terminology corrections via Discussions as a standing request — that is the honest form of review available to the repo today, and it is stated in the document.

Tests-first evidence

tests/test_arabic_readme.py written first: red run 9/9 failed (no README.ar.md). After authoring: 9 passed — pinning the RTL wrapper, two-sided switcher, Arabic boundary + English original, no-review stance + fiqh-basis link, honest result + record link + file existence, verbatim quickstart commands + CDP caveat, the five terms, the scope statement, and the disclaimers.

Gates

• uv run ruff check keel tests packages — All checks passed!
• uv run mypy — Success: no issues found in 235 source files
• uv run pytest -q — 2841 passed, 1 skipped

Scope check

• No rail, rule, or default classification is touched — documentation and pinning tests only.