Allow adding instructions when approving permission prompts (like Claude Code's "Yes, and...")

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

When Copilot CLI presents a permission prompt (e.g., to run a command or edit a file), the options are multiple choice — you can select "Yes", "Yes, always", or "No". If you select "No", you can press Tab to expand and type a reason/alternative instructions. However, there is **no equivalent Tab-to-expand option on "Yes"** to approve the action while also providing additional guidance.

In Claude Code, you can press Tab on the "Yes" option to expand it and add follow-up instructions — effectively a "Yes, and..." that approves the pending action and feeds extra context to the agent in one step.

### Proposed solution

Add Tab-to-expand support on the "Yes" option at permission prompts, matching the existing behavior on "No". When the user presses Tab on "Yes", they can type additional instructions that are passed to the agent alongside the approval.

### Example prompts or workflows

1. Copilot proposes: `rm -rf dist/`
- User selects "Yes", presses Tab, types: "and then rebuild with npm run build"
- Result: command is approved, agent also knows to rebuild after

2. Copilot proposes editing `src/config.ts`
- User selects "Yes", presses Tab, types: "make the same change in config.test.ts too"
- Result: edit is approved, agent picks up the follow-up instruction

### Additional context

The "No" option already supports this Tab-to-expand pattern for providing rejection reasons. This request is to mirror that UX on the approval side, matching Claude Code's behavior. It significantly reduces friction — without it, you have to approve, wait for the agent to continue, and then type a separate follow-up message.

The issue names no files or tests; start by locating the permission prompt implementation and the existing Tab-to-expand handling for the "No" option. Add equivalent behavior for "Yes" so approval and extra instructions are both passed to the agent, then verify the described command and file-edit workflows.

Escrito por el modelo de indexación a partir del texto del issue.