Add ILaaS provider

[lnussbaum]

• Add ILaaS (https://www.ilaas.fr/)
• Start with adding a single model, but others (listed on https://www.ilaas.fr/liste-des-modeles-llms/) will follow in a subsequent PR.

[github-actions]

Action items

• [high] [violation] providers/ilaas/logo.svg - Check: New providers must ship a compliant logo. Why: This PR adds providers/ilaas/provider.toml and a model but the diff contains no providers/ilaas/logo.svg. AGENTS.md makes a compliant logo a hard merge blocker for every new provider, and comparable OpenAI-compatible providers serving GLM-5.2 (tinfoil, ambient) all ship one. Action: Add providers/ilaas/logo.svg using currentColor with no fixed size/hardcoded color and a square viewBox.
• [medium] [possible mistake] providers/ilaas/models/scaleway/glm-5.2.toml:2 - Check: reasoning_options must reflect the inference provider's actual API surface, not the upstream model's. Why: The declared effort set ["none", "low", "medium", "high"] is not evidenced by any ILaaS API reference in the PR, and it differs from other providers serving the same zhipuai/glm-5.2 base model (tinfoil declares 7 effort levels, ambient declares ["high", "max"]). Without ILaaS-specific evidence the values may misrepresent what callers can control. Action: Cite the ILaaS API documentation that confirms these exact effort values (and the request parameter used), or set reasoning_options = [] if no caller control is verified.
• [low] [possible mistake] providers/ilaas/models/scaleway/glm-5.2.toml:6 - Check: Material factual overrides (pricing and limits) need a mapped source citation. Why: cost (input=1.8, output=5.5) and limit overrides (context=256_000, output=16_384) reduce the base model's 1M/131K window and introduce pricing, but the PR body only links the ILaaS homepage and model-list page without stating what either URL supports. Action: Add a direct ILaaS pricing/model-doc citation to the PR body (or a leading comment block in the file) stating which source backs the cost and limit values.