feat: add adaptive AI mentor system

[ishijain2212]

Summary

Adds an adaptive AI mentor system to PyBe that personalizes mentoring based on learner reasoning, misconceptions, relevant learning history, learner profile, and adaptive difficulty.

What changed

• Added scenario-aware reasoning and Python code generation
• Added relevant learner history selection
• Added learner profile generation
• Added adaptive mentoring strategies:
  • REMEDIATE
  • EXPLAIN
  • ENCOURAGE
  • CHALLENGE
• Added adaptive difficulty:
  • FOUNDATION
  • INTERMEDIATE
  • ADVANCED
• Added Gemini-powered mentor responses
• Added structured mentor response validation and fallback handling
• Added structured mentor output to the frontend
• Integrated the complete adaptive pipeline into session processing

Adaptive flow

Learner reasoning
→ misconception/history/profile analysis
→ mentoring strategy
→ adaptive difficulty
→ Gemini mentor
→ personalized next question

Example

For the "Find the Longest Pencil" scenario, the system detected a gap around initializing the current longest value and selected:

• Strategy: REMEDIATE
• Difficulty: FOUNDATION

Gemini then generated a corrective question targeting that specific gap.

Testing

Tested the complete session flow locally with the adaptive mentor and verified scenario-grounded Python generation and personalized mentor output.