chore: refresh seeded AI model catalog (Google/Anthropic/OpenAI)

[esafwan]

What

Refreshes create_demo_ai_models() in huf/install.py against the providers' current model lineups (July 2026) — deprecated and legacy-but-still-available models pruned, OpenRouter open-weight lineup expanded.

Google

• ❌ Removed: gemini-2.0-flash-001, gemini-2.0-flash-lite-preview (deprecated, shut down 2026-06-01), gemini-3-pro-preview (superseded by 3.1)
• ✅ Added: gemini-3.5-flash, gemini-3.1-pro-preview, gemini-3.1-flash-lite, gemini-3-flash-preview, gemini-3.1-flash-image (Nano Banana 2), gemini-embedding-001, gemini-embedding-2
• Kept: gemini-2.5 family, gemma-3, nano-banana-pro, text-embedding-004

Anthropic

• ❌ Removed: claude-2, claude-opus-4.1 (deprecated), plus legacy: claude-sonnet-4.5, claude-opus-4, claude-opus-4.5, claude-3.7-sonnet, claude-3.5-sonnet, claude-3.5-haiku, claude-sonnet-4-20250514
• ✅ Added: claude-fable-5, claude-opus-4.8, claude-sonnet-5 (kept claude-haiku-4.5 — current)

OpenAI

• ❌ Removed: gpt-3.5-turbo, gpt-4-turbo, gpt-4o, gpt-4.1-nano, o1-preview, o1-mini, dall-e-3, gpt-image-1 (deprecated), plus legacy: gpt-4.1, gpt-4.1-mini
• ✅ Added: gpt-5.6-sol / gpt-5.6-terra / gpt-5.6-luna, gpt-5.5, gpt-5.4 / gpt-5.4-mini / gpt-5.4-nano, gpt-5.2, gpt-image-2
• Kept: gpt-5/5.1/5-mini/5-nano, gpt-4o-mini (not deprecated), whisper-1, embeddings

OpenRouter

• ❌ Removed: google/gemini-2.0-flash-exp:free, openai/gpt-4.1-mini, openai/gpt-4.1-nano, anthropic/claude-4.5-sonnet-20250929
• ✅ Added all 14 current :free tier models, verified live against openrouter.ai/api/v1/models (2026-07-20): gemma-4-31b/26b, gpt-oss-20b, nemotron-3 x7, cohere north-mini-code, poolside laguna x2, tencent hy3. The :free suffix stays in model_name — it is both the free label and the exact id litellm routes with.
• 🔧 Fixed stale ids that no longer exist on OpenRouter: qwen/qwen3-coder:free → qwen/qwen3-coder, google/gemma-3-27b-it:free → google/gemma-3-27b-it
• ✅ Added (current open-weight flagships, per OpenRouter catalog July 2026): deepseek/deepseek-v4-pro + deepseek-v4-flash, z-ai/glm-5.2, moonshotai/kimi-k3 + kimi-k2.7, minimax/minimax-m3 + m2.7, microsoft/phi-4, google/gemma-4-31b-it, qwen/qwen3.7-max + qwen3.6-plus
• Kept: openai/gpt-5 series, gpt-4o-mini, google/gemini-2.5-flash + gemma-3-27b:free, deepseek-chat-v3-0324/v3.1, qwen3-vl-235b + qwen3-coder:free, minimax-m2

Notes

• Seed-only change (new installs); existing sites keep their AI Model rows — the seeder only inserts when the model doesn't exist.
• Sources: Google Gemini API docs, Anthropic models overview, OpenAI models page, OpenRouter catalog references (July 2026); GPT-5.6 tiers confirmed via OpenAI announcement.