Show timestamps next to messages in conversation view

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

When returning to a long-running Copilot CLI session after a break (e.g., coming back to the terminal after a few hours), there's no way to see **when** each message was sent. The conversation reads as a flat transcript with no time context.

This makes it hard to:
- Know when you last interacted with the session
- Understand the pacing of a conversation (was that exchange 5 minutes ago or 5 hours ago?)
- Orient yourself when resuming after a gap

The agent already receives `` with each user message, so the data exists — it's just not displayed.

### Proposed solution

Show a **timestamp** next to each user and assistant message in the terminal UI. Something like:

```
[07:23] > How does the G502 compare to the G502 X?

[07:23] The G502 X is essentially a modernised rebuild...

[13:42] > I just came back — when was my last message?
```

Options to consider:
- **Relative timestamps** for recent messages ("2m ago"), absolute for older ones ("07:23")
- **Gap indicators** — when there's a long gap between messages (e.g., >30 min), show a separator: `--- 6 hours later ---`
- **Configurable** — a setting in config.json to enable/disable, or choose format (relative / absolute / off)

### Example prompts or workflows

User works in a session in the morning, walks away, comes back in the afternoon. Currently they see a flat wall of text with no temporal context. With this feature they'd immediately see when the last exchange happened and how long they've been away.

### Additional context

The `` tag is already injected into each user message for the agent's benefit — this is about surfacing that same information in the UI for the human.

Start by locating the terminal conversation view and tracing how the existing value is injected into user messages. Determine how timestamps should be presented for user and assistant messages, including whether config.json should control the format or visibility. Done means the conversation shows useful time context without disrupting the existing transcript.

由索引模型根据 Issue 内容生成。