[Bug] AI Bot Produces Irrelevant / Non-Debate Responses During Opening Statements

### Description
During a **User vs Bot debate**, the AI bot generates responses that are **off-topic, unserious, and not aligned with debate rules**, especially during the **Opening Statements** phase.

Instead of presenting a structured argument relevant to the debate topic, the bot outputs casual filler text (e.g., surfer-style or idle chatter), which breaks the core functionality and credibility of the debate system.

---

### Steps to Reproduce
1. Start the backend and frontend locally.
2. Log in successfully.
3. Create or join a **User vs Bot** debate.
4. Select any topic (e.g., *“Should AI rule the world?”*).
5. Reach the **Opening Statements** phase.
6. Wait for the bot to generate its opening statement.

---

### Expected Behavior
- The AI bot should:
- Respect the debate **topic**, **phase**, and **assigned stance**.
- Produce a **clear, structured argumentative response**.
- Present logical reasoning suitable for a competitive debate.
- Maintain relevance even if the persona tone is casual.

**Example (Expected):**
> “AI governance could reduce human bias, improve efficiency, and make data-driven decisions at scale, which is why…”

---

### Actual Behavior
- The bot responds with **irrelevant, casual, or nonsensical text** that:
- Does not address the topic.
- Lacks an argument or stance.
- Sounds like general chatbot chatter.

**Example (Actual):**
> “No way, man! Dude, I’m spaced out, man! Chill, I’ll catch the next wave at the beach diner.”

---

### Screenshots / Evidence

---

### Root Cause (Analysis)
- The AI prompt sent from the backend:
- Does **not enforce debate constraints** (topic, phase, stance).
- Allows persona prompts to override debate rules.
- Lacks system-level instructions for argumentative structure.
- No output validation exists to reject low-quality or irrelevant AI responses.

---

### Impact
- Breaks the core purpose of DebateAI.
- Reduces user trust in AI debates.
- Makes bot debates unusable for serious discussion, demos, or evaluation.
- Affects judging, scoring, and learning outcomes.

---

### Suggested Fix
1. Strengthen the **system prompt** for AI debaters:
- Explicitly include:
- Debate topic
- Phase (Opening / Rebuttal / Closing)
- Assigned stance
- Requirement for structured arguments
2. Ensure persona affects **tone only**, not content quality.
3. Add backend validation to reject:
- Off-topic responses
- Extremely short or non-argumentative outputs
4. Provide a fallback regeneration if validation fails.

---

### Environment
- OS: Windows 10
- Backend: Go (Gin)
- Frontend: React + Vite
- MongoDB: Local instance
- Debate Mode: User vs Bot
- Phase: Opening Statements

---

### Additional Context
This issue is **not related to authentication, WebSockets, or UI rendering**.
The debate state and frontend display are correct; the failure originates from AI prompt design and response handling.

---

### Priority
**High** – Core feature malfunction

---

### Willingness to Contribute
I am willing to:
- Improve the AI system prompt
- Add validation logic
- Submit a PR with fixes if guided on preferred design direction