Feature request: add auto reasoning_effort option

Codex should support an auto option for reasoning_effort, ideally through the API as well.

users should not have to think about how much thinking the model needs for every prompt. having to manually choose low, medium, high, or xhigh makes us worry more about usage limits and speed instead of focusing on the actual task.

Anthropic already has a similar idea with Claude’s adaptive thinking, where the model can decide whether and how much to use extended thinking based on task complexity. OpenAI should have an equivalent option for reasoning effort.

from the user side, i just want an option like:

{
  "reasoning_effort": "auto"
}

this would make Codex, ChatGPT, and API usage easier and more efficient: simple tasks stay cheap and fast, while harder tasks can automatically use more reasoning when needed.