docs: Add comprehensive architecture review and refactoring plan (v6.0.0)

[SatoryKono]

Summary

Add a comprehensive architecture review document (ARCHITECTURE_REVIEW_2026-03-04.md) that evaluates the BioETL project across 10 key categories using a weighted scoring system. The review provides detailed analysis of current state, identifies 8 major issues, and proposes a prioritized 5-phase refactoring roadmap with concrete metrics and success criteria.

Changes

• Added docs/00-project/ARCHITECTURE_REVIEW_2026-03-04.md (587 lines)
  • Numerical assessment across 10 categories (weighted score: 8.20/10)
  • Detailed evaluation of architecture layers, modularity, DDD implementation, testing, error handling, observability, performance, security, documentation, and technical debt
  • 8 prioritized problems (P-001 through P-008) with impact analysis
  • 8 refactoring initiatives (RF-001 through RF-008) with concrete implementation steps
  • 5-phase roadmap with projected score improvements (8.20 → 8.77)
  • Recommended metrics and new tests for quality control
  • Mapping of metrics to integral score calculation

Type

• Documentation

Affected layers

• Domain
• Application
• Infrastructure
• Composition
• Interfaces
• Configs

Test plan

N/A — Documentation only, no code changes.

Checklist

• No new import boundary violations (documentation only)
• No hardcoded secrets (documentation only)
• Type annotations not applicable (documentation only)
• No code changes requiring tests

Notes

This architecture review serves as a baseline assessment and strategic planning document for the BioETL v6.0.0 release. Key findings:

Strengths:

• Excellent layered architecture (9.0/10) with zero import violations
• Comprehensive testing infrastructure (9.0/10) with 634 architecture tests
• Strong DDD implementation (8.5/10) with 52 port protocols
• Strict type safety (mypy --strict) with 25+ CI workflows

Priority improvements:

1. RF-001: Decompose DataNormalizationService (36 methods → specialized services)
2. RF-002: Reduce Any usage from 385 to <200 instances
3. RF-003: Break down 32 files >500 LOC (largest: 952 lines)
4. RF-004: Unify transformer logic across 6 provider implementations

The document provides actionable guidance for the development team to incrementally improve code quality while maintaining architectural discipline.

https://claude.ai/code/session_01XtxheHFMpq5RCAexnNxc56

[SatoryKono]

Closing: stale docs/reports/audit branch, content outdated.