Support COPILOT_MODEL_EFFORT for reasoning effort level

### Describe the feature or problem you'd like to solve

GitHub Copilot CLI lacks a mechanism to adjust "reasoning effort" for reasoning models (such as the gpt-5 series) when using BYOK

### Proposed solution

When a user sets `COPILOT_MODEL_EFFORT` environment variable, the CLI should pass this parameter to the underlying model API.

This feature will enable developers to use a lower reasoning level through environment variable and to increase the reasoning level for complex architectural designs or debugging tasks to obtain more accurate output, thereby enhancing the utility of the CLI tool in professional development scenarios.

### Example prompts or workflows

```powershell
$env:COPILOT_PROVIDER_TYPE='azure'
$env:COPILOT_PROVIDER_WIRE_API='responses'
$env:COPILOT_PROVIDER_BASE_URL='https://resource-name.openai.azure.com'
$env:COPILOT_PROVIDER_API_KEY=$env:AOAI_API_KEY
$env:COPILOT_MODEL_EFFORT='xhigh' # <-- HERE
$env:COPILOT_MODEL='gpt-5.4'
copilot
```

### Additional context

N/A

Start by tracing how COPILOT_MODEL and provider settings reach the underlying model API. Add support for reading COPILOT_MODEL_EFFORT and passing its value for reasoning models, then verify that the example Azure Responses workflow sends the requested effort level.

由索引模型根据 Issue 内容生成。