feat: PIXIE adaptive hint ladder driven by playerProfiles

[ibloud]

Summary

Implement an adaptive PIXIE hint ladder that personalizes hint depth and pacing using playerProfiles traits and recent performance signals.

Why

Hints should meet players at their current skill/learning state rather than being one-size-fits-all.

Scope

• Define hint ladder tiers (e.g. nudge → pattern cue → tactical line → concrete move).
• Map profile traits/performance to starting tier and escalation pace.
• Add guardrails to prevent over-assistance and preserve challenge.
• Log hint interactions for learning loop:
  • tier shown, accepted/ignored, follow-up outcome

Integration

• Read minimal profile slice required for hint decisions.
• Fallback behavior when profile missing or low confidence.
• Privacy-conscious handling of profile-derived personalization.

Acceptance Criteria

• Hint tier selection changes based on profile/performance inputs.
• Escalation is deterministic and test-covered.
• Telemetry is captured for post-launch tuning.
• Safe fallback works when profile data unavailable.