Bug: `explore` tool hardcodes model to `gpt-5.4-mini`, ignoring custom/DeepSeek API configuration

### Describe the bug

After updating to the latest version of the GitHub Copilot CLI (v1.0.65), agent decision/attempt to invoke the `explore` tool ignores the configured custom model (e.g., DeepSeek endpoint configuration) and instead attempts to pass `gpt-5.4-mini` to the API endpoint. This causes upstream API providers like DeepSeek to fail with a `400 Bad Request` because the model parameter is invalid for their endpoint.

```
✗ Explore(gpt-5.4-mini) Explore tracker button color flow
└ No response generated

✗ 400 {"error":{"message":"The supported API model names are deepseek-v4-pro or deepseek-v4-flash, but you passed gpt-5.4-mini.","type":"invalid_request_error","param":null,"code":"invalid_request_error"}}
```

### Affected version

GitHub Copilot CLI 1.0.65

### Steps to reproduce the behavior

1. Configure GitHub Copilot CLI to use an external custom endpoint / DeepSeek API.
3. Run any command using the autonomous explore tool feature.
5. See error.

### Expected behavior

The `explore` tool should respect the local configuration or environment variables defining the target model (e.g., `deepseek-v4-pro`), rather than falling back to a hardcoded default model name string during the tool call payload execution.

### Additional context

- **OS:** Windows 11 / WSL2 (Ubuntu 24.04 LTS)

Start by reproducing the explore tool call with the external DeepSeek endpoint configuration described in the issue. Trace the explore tool's model selection and request payload, then verify that the configured model is forwarded instead of gpt-5.4-mini and that the request succeeds against the configured endpoint.

Do mô hình lập chỉ mục viết ra từ nội dung của issue.