[opencode/openai part 1] Add reasoning options

[rekram1-node]

Adds OpenCode Zen reasoning-effort metadata for 15 OpenAI models.

Provider behavior

• Zen routes these model IDs through its Responses endpoint using @ai-sdk/openai: https://opencode.ai/docs/zen/
• OpenCode configures this as the AI SDK reasoningEffort option: https://opencode.ai/docs/models/
• The Responses wire field is reasoning.effort, and accepted values are model-dependent: https://platform.openai.com/docs/guides/reasoning

Model evidence

• GPT-5.1 supports none, low, medium, high: https://platform.openai.com/docs/models/gpt-5.1
• GPT-5.2 supports none, low, medium, high, xhigh: https://platform.openai.com/docs/models/gpt-5.2
• GPT-5.3-Codex supports only low, medium, high, xhigh: https://platform.openai.com/docs/models/gpt-5.3-codex
• OpenAI Codex canonical metadata enumerates the same four GPT-5.3-Codex levels: https://github.com/openai/codex/blob/main/codex-rs/models-manager/models.json#L291-L312
• GPT-5.4, Mini, and Nano support none, low, medium, high, xhigh: https://platform.openai.com/docs/models/gpt-5.4 https://platform.openai.com/docs/models/gpt-5.4-mini https://platform.openai.com/docs/models/gpt-5.4-nano
• GPT-5.4 Pro and GPT-5.5 Pro support medium, high, xhigh: https://platform.openai.com/docs/models/gpt-5.4-pro https://platform.openai.com/docs/models/gpt-5.5-pro

No token-budget or binary-toggle controls are claimed. Scope is limited to opencode/openai and supersedes that portion of #2247.