[FEATURE]: Add real-time token usage visibility via API response and backend logging

### Feature and its Use Cases

## Problem

The application currently lacks built-in visibility into token consumption during AI debates. This makes it difficult for developers to:

- Monitor API usage and cost in real-time
- Debug prompt efficiency and response overhead
- Understand token usage per debate turn without external tools

At present, token usage can only be inferred indirectly (e.g., dashboards or database inspection), which is not convenient during active development or testing.

---

## Proposed Solution

Introduce lightweight token observability directly within existing system flows, without requiring additional tools.

### 1. API Response Enhancement
Include token usage metadata in the JSON response for:
- Debate message responses
- Judge responses

Fields:
- `prompt_tokens`
- `response_tokens`
- `total_tokens`

This allows developers to inspect token usage directly via the browser Network tab (F12) without modifying the frontend UI.

---

### 2. Backend Logging
Add structured logging in the backend controller to print token usage for each AI interaction.

Example:

[TOKEN USAGE] Bot: Yoda | Prompt: 420 | Response: 69 | Total: 489

This enables real-time monitoring directly from the server terminal.

---

## Expected Outcome

- Token usage is visible instantly via:
- Browser Network tab (API response)
- Backend server logs
- No additional setup or tools required
- No impact on existing frontend behavior
- Improved debugging, observability, and cost awareness

---

## Benefits

- Zero-friction developer experience
- Immediate visibility into AI token consumption
- Helps optimize prompts and model performance
- Lays groundwork for future features like token analytics dashboards

---

## Compatibility

- No breaking changes to existing frontend
- Token fields are additive and optional

### Additional Context

## Additional Context

- Token usage data is already available from the Gemini API via `usageMetadata`, but is not currently exposed to developers during runtime.
- This proposal focuses on surfacing that existing data with minimal changes and zero impact on current user flows.
- The approach avoids introducing new dependencies or tools, keeping the implementation lightweight and maintainable.
- This can serve as a foundation for future enhancements such as token usage dashboards, cost estimation, or analytics features.

### Code of Conduct

- [x] I have joined the [Discord server](https://discord.gg/hjUhu33uAn) and will post updates there
- [x] I have searched existing issues to avoid duplicates