feat: Adaptive Learning System - Smart Mastered Problem Resurfacing

[smithrashell]

Problem Statement

Problems at box level 6-8 (mastered) are reviewed based on simple time intervals (60-120 days). This is inefficient because:

1. Easy reviews don't strengthen memory (no "desirable difficulty")
2. Review time could be spent on new learning
3. Mastered problems should resurface contextually, not just on schedule

Proposed Solution: Triggered Reviews

Instead of showing mastered problems on a fixed schedule, resurface them when contextually relevant:

1. Failure Trigger: User fails Problem B → find related mastered Problem A → show A to reinforce foundation
2. Chronic Struggle Trigger: Problems with 3+ unsuccessful attempts → find related mastered problems to reinforce concepts

Session Composition Priority

1. Triggered reviews (max 2 mastered problems related to struggling problems)
2. Learning reviews (box 1-5, still learning)
3. New problems (primary learning)
4. Passive mastered reviews (box 6-8, only if session not full)

Smart Selection Algorithm

• Find "bridge" problems with high aggregate relationship to ALL struggling problems
• Scoring: aggregateStrength * (1 + coverageBonus)
• Relationship threshold: >= 2.0

Tag Mastery Decay Integration

• If a tag's decay_score < 0.7, boost priority for triggered reviews with that tag

Implementation Tasks

• Add getTriggeredReviews and getProblemsNeedingReinforcement to problem_relationships.js
• Add getProblemsWithHighFailures query to problems.js
• Integrate triggered reviews into problemServiceSession.js
• Create unit tests for triggered review detection

Expected Outcomes

Metric	Before	After
Mastered reviews per session	1-2 (scheduled)	0-2 (contextual)
New problems per session	2-3	3-4
Review effectiveness	Low (easy)	High (reinforcing weakness)

[smithrashell]

Completed in 385695c and 292789d. Adaptive triggered reviews with failure-based bridge problems and prerequisite detection are implemented and merged to main.