feat: implement scoring/ — SCI scoring engine (Front 1)

[guidevit-dealsmartai]

Context

Part of the Application Layer design (Issue #7). Front 1 of 4 — merges first.

Scope

Implement the SCI (Score de Consistência) scoring engine under scoring/. This module reads from PostgreSQL and Neo4j (populated by the data pipeline), calculates a 0–1000 consistency score across 6 dimensions, and writes results back to the database.

Deliverables

Dimension Calculators (scoring/dimensions/)

Dimension	File	Weight	Max Deduction
Wealth evolution vs salary	patrimonio.py	25%	−250 pts
Vote × donor sector correlation	voto_doador.py	20%	−200 pts
Family company contracts	contrato_familiar.py	20%	−200 pts
Amendment beneficiary linkage	emenda_vinculo.py	15%	−150 pts
Family office hiring	gabinete_familiar.py	12%	−120 pts
Vote change after donation	voto_pos_doacao.py	8%	−80 pts

Each dimension class inherits from BaseDimension and implements calculate(mandatario_id) → DimensionResult(score, deductions, inconsistencies).

Core Modules

• engine.py — SCIEngine.calculate(mandatario_id) → SCIResult — orchestrates all 6 dimensions, applies weights, returns 0–1000 score
• detector.py — InconsistencyDetector — creates :Inconsistencia records in PG + Neo4j with FATO/MÉTRICA/FONTE
• history.py — ScoreHistory — records score snapshots for audit trail
• queries/neo4j_queries.py — Cypher queries for graph traversals (family→company→contract)
• queries/pg_queries.py — SQL aggregations (wealth, votes, donations)

CLI

python -m scoring.engine {mandatario_id | --all} — can be invoked by Airflow DAG or manually.

Key Constraints

• Fixed weights (no ML — that's Phase 2)
• Score = 1000 − sum(deductions), clamped to [0, 1000]
• Every deduction MUST produce an :Inconsistencia with FATO + MÉTRICA + FONTE
• All scores deterministic and reproducible
• Imports vigiabr-schemas from pipeline/schemas/

Dependencies

• Depends on: pipeline/schemas/ (merged), data in PostgreSQL + Neo4j
• Blocks: Backend API (Front 2) reads pre-computed scores

Branch

feat/8-sci-scoring-engine (from develop)