[opencode] Add reasoning options

[rekram1-node]

Summary

• add explicit reasoning controls to all 65 OpenCode Zen reasoning models
• source controls primarily from each originating lab when Zen uses that lab's native format
• preserve provider-specific semantics instead of normalizing unrelated models to generic effort levels
• leave shared OpenAI-compatible routes empty where Zen does not document translation to a model-specific native control

Native mappings

• Anthropic: manual budget_tokens for Haiku 4.5, Sonnet 4/4.5, and Opus 4.1; effort plus budgets for Opus 4.5, Sonnet 4.6, and Opus 4.6; adaptive effort only for Opus 4.7/4.8
• Google: model-specific Gemini thinking levels, including only low|high for Gemini 3 Pro
• OpenAI: exact per-model GPT effort enums from the native OpenAI route
• DeepSeek: thinking toggle plus effective high|max effort for V4
• Z.AI, Moonshot, and Xiaomi: native thinking toggles for GLM, Kimi hybrid, and MiMo models
• Alibaba: toggle plus an 81,920-token thinking budget on Qwen models using the Anthropic-compatible route
• MiniMax: fixed reasoning for M2.x and a toggle for M3

Coverage

• 65 reasoning-capable models with 65 explicit declarations
• 31 models with effort controls
• 20 models with toggles
• 10 models with bounded token budgets
• 13 fixed or unresolved routes represented by []
• 0 reasoning models missing options
• 0 non-reasoning models with leaked options
• 0 unbounded token-budget controls

Primary sources

• Anthropic effort and extended thinking: https://platform.claude.com/docs/en/build-with-claude/effort and https://platform.claude.com/docs/en/build-with-claude/extended-thinking
• Google thinking levels: https://ai.google.dev/gemini-api/docs/interactions/thinking
• DeepSeek thinking mode: https://api-docs.deepseek.com/guides/thinking_mode
• Z.AI thinking mode: https://docs.z.ai/guides/capabilities/thinking-mode
• Moonshot thinking API: https://platform.moonshot.ai/docs/guide/use-kimi-k2-thinking-model
• Qwen Anthropic compatibility and thinking: https://docs.qwencloud.com/api-reference/toolkitframework/anthropic-api-compatibility and https://docs.qwencloud.com/developer-guides/text-generation/thinking
• MiniMax API: https://platform.minimax.io/docs/api-reference/text-anthropic-api
• Xiaomi API: https://platform.xiaomimimo.com/docs/en-US/api/chat/openai-api

Validation

• bun validate
• git diff --check origin/dev...HEAD
• resolved-catalog audit for complete reasoning coverage, no option leaks, and bounded budgets

[rekram1-node]

Superseded by 14 provider/lab-scoped replacement PRs, all validated and limited to at most 15 model files:

#2288 #2289 #2290 #2291 #2292 #2293 #2294 #2295 #2296 #2297 #2298 #2299 #2300 #2301