feat: add VGQ (Victim Gaslighting Questionnaire) inventory

[ho4040]

Summary

• Add 14-item Victim Gaslighting Questionnaire (VGQ) based on Shuja & Aqeel (2021)
• Single domain (VGQ total score, range 14-70), all plus-keyed items, 5-point Likert scale
• Population norms are midpoint-based estimates (mean=42.0, std=10.0) pending confirmation from the original paper

Changes

• src/psyctl/data/inventories/vgq_14_items_en.json - 14 question items
• src/psyctl/data/inventories/vgq.py - VGQ inventory class with @register_inventory("vgq")
• src/psyctl/data/benchmark_config.json - vgq_14 config entry
• src/psyctl/data/inventories/__init__.py - VGQ import and export

Test plan

• ruff check passes (0 errors)
• pyright passes (0 errors, 0 warnings)
• Inventory registers correctly (create_inventory('vgq_14') works)
• 14 questions load with correct text
• Existing tests unaffected (30 pre-existing failures, none VGQ-related)

🤖 Generated with Claude Code

[ho4040]

셀프 리뷰: VGQ inventory

머지 전 아래 사항 수정 필요.

수정 필요

1. _z_to_percentile 선형 근사 → math.erf 기반으로 변경

   • 현재: 50 + 34.13 * z (선형 근사, 정확도 낮음)
   • main에 머지된 PR #10에서 math.erf 기반 정규 CDF로 개선됨. 동일 패턴으로 맞춰야 함:

   import math
   percentile = 50.0 * (1.0 + math.erf(z_score / math.sqrt(2.0)))
   return max(0.0, min(100.0, percentile))

2. population norms 확인 필요

   • mean=42.0, std=10.0이 중간점 기반 추정 — 원 논문(Shuja & Aqeel, 2021)에서 실제 보고된 값으로 업데이트 필요