feat(ofox): add Ofox provider (13 top-tier models)

[celeste1900]

What

Adds Ofox (https://ofox.ai) as a provider with 13 top-tier models.

Ofox is an LLM gateway: one API key gives access to models across providers via three natively-supported protocols — OpenAI (/v1), Anthropic (/anthropic), and Gemini (/gemini).

Models (13)

Frontier: gpt-5.5, gpt-5.6-luna, gpt-5.6-sol, gpt-5.6-terra, claude-opus-4.8, claude-sonnet-5, claude-fable-5, gemini-3.1-pro-preview, grok-4.3
Chinese labs: deepseek-v4-pro, qwen3.7-max, kimi-k2.6, glm-5.2

Notes

• All models use base_model inheritance; logo uses currentColor; bun run validate passes.
• Costs are taken from Ofox's public catalog (GET https://api.ofox.ai/v1/models).
• Chinese-lab pricing: Ofox connects to these providers' domestic (RMB-billed) channels, so USD figures are RMB→USD conversions and may differ slightly from a provider's own USD international list (e.g. deepseek-v4-pro is $0.45/$0.88 here vs DeepSeek's USD list $0.435/$0.87 — a ~3% RMB-conversion difference, not a markup).
• gemini-3.1-pro-preview intentionally has no cache_write: Gemini bills cache as an hourly storage cost, not a per-write charge, so there's no per-token write price to list.
• reasoning_options left empty pending per-protocol verification of which reasoning controls Ofox forwards.
• Supersedes feat(ofox): add Ofox provider (curated top-tier models) #3253 (closed); minimax-m3 dropped pending a context-tier pricing clarification and will follow in a later PR.

[github-actions]

No actionable findings.

[celeste1900]

Ready for review whenever a maintainer has a moment — validate and the review bot are both green (no actionable findings), and it's mergeable/clean. Small curated set of 13 models using base_model inheritance + currentColor logo. Happy to adjust if anything's needed.