Effort mismatch: CLI launch effort differs from effort shown in VS Code chat UI

### Describe the bug

When launching Copilot CLI with an explicit --effort value (for example xhigh), the generated VS Code chat shows a different effort (for example Medium).

The core issue is an effort mismatch between the CLI launch configuration and the VS Code chat UI display.
This is reproducible with models other than gpt-5.3-codex as well.

### Affected version

GitHub Copilot CLI 1.0.63

### Steps to reproduce the behavior

1. Run Copilot CLI with an explicit effort value:
copilot --model gpt-5.3-codex --effort xhigh --context long_context --log-level debug --log-dir /tmp/mylog -p "test"

2. Check logs:
grep "defaultReasoningEffort" /tmp/mylog/*.log

3. Open the generated VS Code chat and check the effort shown in the chat UI footer.

**Actual Result**
- Logs indicate the requested effort was set (example: defaultReasoningEffort=xhigh).
- The VS Code chat UI shows a different effort level (example: Medium).

### Expected behavior

The effort shown in the generated VS Code chat UI should match the effort specified at CLI launch (--effort), or the UI should clearly indicate if it represents a different metric/concept.

### Additional context

- This issue is not model-specific.
- Reproduced across multiple models; the mismatch is between launch effort and effort displayed in VS Code chat.

Run the reported Copilot CLI command with an explicit --effort value, inspect defaultReasoningEffort in the generated logs, and compare it with the effort shown in the generated VS Code chat footer. Trace how the launch effort is transferred to the chat and verify that the displayed value matches, or that the UI clearly distinguishes a different metric.

由索引模型根據 Issue 內容生成。