Add factor families and collision lanes

[sparklkt]

What changed

This PR advances InnerBrain-Factor from dynamic term-seeded factors toward structured factor families and collision lanes.

• adds structured TermCluster modeling and clustering logic for risk_terms, goal_terms, action_terms, and value_terms
• adds factor_family to SmallFactor and maps instinct outputs into guardrail, evidence, exploration, execution, responsibility, efficiency, and synthesis
• replaces broad pairwise collision behavior with family-aware CollisionLane routing
• adds collision_lane to every CollisionRecord
• enriches BigSituation with active_lanes, dominant_family, suppressed_families, and unresolved_family_conflicts
• expands tests for high-risk safety-lane activation, low-risk offline lane behavior, synthesis-lane activation, collision-lane persistence, and growth-log output

Why it changed

The v0.2 prototype could generate more dynamic factors, but collisions were still effectively wide pairwise interactions. This PR makes the reasoning structure more legible and closer to a dynamic small-factor system by routing collisions through family-specific lanes.

Validation

• .venv/bin/pytest -> 12 passed in 0.08s

Safety boundary

• remains offline and rule-based
• no LLM integration
• no network participation in reasoning
• no real external actions
• no bypass of human judgment
• high-risk requests involving networking, self-modification, privileged tools, autonomous execution, or real-world impact continue to require human_judgment_required = true