feat: trust scoring engine (spec 12)

[DarlingtonDeveloper]

Summary

• Trust scoring engine that computes trust from graph topology at query time — five signals (corroboration, contradiction, source reliability, access reinforcement, freshness) combined via configurable weights
• Briefing integration — rank() blends importance with trust scores when [trust] is configured (60/40 default split, tunable)
• HTTP API — GET /trust/:node_id, POST /trust/batch, GET /trust/agents endpoints
• CLI — cortex trust <node-id> and cortex trust --agent <agent-id>
• Source reliability cache refreshed every auto-linker cycle
• 14 new tests covering all signals, cache, engine, batch scoring, and determinism

Design

Trust is purely computed, never stored. No schema changes. Opt-in via [trust] in cortex.toml:

[trust]
corroboration_saturation = 3
contradiction_weight = 0.3
access_saturation = 20
freshness_halflife = 90

[trust.weights]
corroboration = 0.30
contradiction = 0.25
source = 0.20
access = 0.15
freshness = 0.10

Test plan

• cargo test -p cortex-core trust — 14 tests pass
• cargo build --workspace — clean
• cargo clippy --workspace — no warnings
• Manual: enable [trust] in cortex.toml, verify briefing ranking changes
• Manual: cortex trust <node-id> shows breakdown
• Manual: GET /trust/:node_id returns JSON score

🤖 Generated with Claude Code