Add user-controlled Sol/Spark model routing

[blackjose007-stack]

Summary

• add two-axis reasoning/model routing with Sol as controller and Spark as bounded coding worker
• add user-owned OFF/TRIAL/ON delegation switch, defaulting to OFF
• record model roles and per-execution-mode duration, retry, token, fallback, and escaped-defect evidence
• document eligible scenarios, prohibited decisions, fallback behavior, and A/B promotion thresholds
• add a tested OFF enforcement gate so worker routes cannot run without user enablement

Validation

• python -m unittest discover -s tests -v (5 PASS)
• skill-creator quick_validate.py (PASS)

Claim boundary

This adds routing policy, persisted user control, and measurement. It does not claim measured token savings yet; Spark remains in TRIAL until comparable observations exist.