feat(analysis): add anomaly detection module (Benford, HHI, round values)

[guidevit-dealsmartai]

Summary

Bring statistical anomaly detection functions from CEAP-Playbook into VigiaBR's processing pipeline.

Motivation

VigiaBR has robust data collection infrastructure (7 sources, Scrapy spiders, Airflow DAGs) but zero anomaly detection code. The CEAP-Playbook repository has battle-tested statistical functions that detect expense anomalies:

• Benford's Law — chi-squared test on first-digit distribution
• HHI (Herfindahl-Hirschman Index) — supplier concentration measurement
• Round values detection — flags suspiciously round amounts

These are exactly the building blocks needed to populate VigiaBR's Inconsistencia model and begin SCI scoring.

Current State

• pipeline/schemas/models/inconsistencia.py — model exists but nothing generates records
• pipeline/processing/processing/transformers/camara.py — skips despesa records (_SKIPPED_TYPES)
• No analysis/detection module exists in the pipeline

Acceptance Criteria

• Analysis module created in pipeline/processing/processing/analyzers/
• Benford, HHI, and round values functions adapted from CEAP-Playbook
• Camara transformer processes expense records (remove from _SKIPPED_TYPES)
• Analyzers generate InconsistenciaCreateSchema records when anomalies detected
• Unit tests for all analyzer functions
• Constants/thresholds configurable

References

• Source repo: https://github.com/jpvss/CEAP-Playbook
• Key files: utils/ceap_utils.py, utils/constantes.py