[anyapi] Add reasoning options

[rekram1-node]

Summary

• backfill reasoning_options for all 23 existing AnyAPI reasoning models
• expose only the common effective low, medium, and high efforts on 8 OpenAI models and Claude Opus 4.6/4.7 plus Sonnet 4.6
• mark the remaining 12 reasoning models fixed/unresolved with []
• avoid inferred toggles, unsupported extra effort levels, and token budgets without exact finite bounds

Audit

• 30 existing AnyAPI models audited
• 23 reasoning models: 11 effort-controlled, 12 fixed/unresolved
• 7 non-reasoning models unchanged
• 0 toggles, 0 token budgets, 0 non-reasoning leaks
• model-only scope: no catalog, base_model, schema, or test changes

Sources

• AnyAPI parameter contract: https://docs.anyapi.ai/guides/parameters
• AnyAPI reasoning-capable backend routing: https://docs.anyapi.ai/features/tag-routing
• AnyAPI exact model catalog: https://anyapi.ai/ai-models
• OpenAI model-dependent reasoning efforts: https://platform.openai.com/docs/guides/reasoning
• Anthropic exact effort support: https://platform.claude.com/docs/en/build-with-claude/effort
• AnyAPI Grok 4.3 fixed reasoning: https://anyapi.ai/ai-models/xai-grok-4-3

Validation

• bun validate
• generated invariant scan: 30 total, 23 reasoning, 11 effort, 12 fixed, 0 missing, 0 non-reasoning leaks, 0 invalid options
• git diff --check