Copilot CLI interactive mode turns blank after submitting a prompt in Windows Terminal, while -p mode works

### Describe the bug

### Summary

In interactive mode, GitHub Copilot CLI renders normally at first, but after I submit a prompt the UI goes blank and no further output is visible.

This happens even with very short prompts that should return only a single sentence.

Non-interactive prompt mode with -p works normally, so the issue appears to be specific to the interactive terminal UI.

### Environment

- GitHub Copilot CLI version: 1.0.71
- OS: Windows
- Terminal: Windows Terminal
- GitHub product version: latest

### Affected version

GitHub Copilot CLI 1.0.71.

### Steps to reproduce the behavior

1. Open Windows Terminal.
2. Start Copilot CLI: copilot
3. Enter any prompt, including a short prompt such as: Reply with one short sentence only.
4. Observe that Thinking appears briefly.
5. The terminal UI then goes blank.

### Expected behavior

After submitting a prompt in interactive mode, Copilot CLI should continue showing Thinking and then render the response in the terminal.

### What actually happened

- Copilot CLI starts and renders normally.
- I enter a prompt.
- It shows Thinking.
- Then the screen turns blank.
- After the blank screen appears, I do not see the response or UI content anymore.

This also happens with very short prompts, including prompts that should only produce a one-sentence response.

### Additional context

I tried the following and still saw the same blank-screen behavior in interactive mode:

- Pressed Ctrl + L
- Pressed ?
- Entered /help
- Tried very short prompts with minimal output
- Tried --screen-reader mode: copilot --screen-reader
--screen-reader mode shows the same issue.

What does work

Non-interactive mode works correctly:

copilot -p "Say hello"

This returns output normally.

Start by reproducing the issue with `copilot` in Windows Terminal using a short prompt, then compare interactive mode with `copilot -p "Say hello"` and `copilot --screen-reader`. The fix is complete when interactive mode continues showing Thinking and renders the response instead of turning blank, while non-interactive mode remains working.

Written by the indexing model from the issue text.