GPT-6 Sol and Luna budgets (openai + azure_openai) #649

Problem

OpenAI released GPT-6 Sol and GPT-6 Luna on 2026-09-22 (API IDs gpt-6-sol,
gpt-6-luna), and both are live on Azure (Foundry catalog + the 2026-09-22
"GPT-6 Astra, Sol, and Luna for production AI agents" announcement). The
bundled openai and azure_openai registries stop at gpt-6-astra, so
scans running on either new model fall back to the 128K default budget and
risk truncated or over-allocated token windows.

Proposed scope

Registry data only, no code or default-model changes:

• gpt-6-sol: 1,050,000 context / 128,000 max output
• gpt-6-luna: 1,050,000 context / 128,000 max output
• Both registries (openai + azure_openai); Azure serves the same stock
  IDs, so the same budgets apply when the deployment name matches.

Verified 2026-09-28 against developers.openai.com/api/docs/models/gpt-6-sol,
developers.openai.com/api/docs/models/gpt-6-luna, and
openai.com/index/introducing-gpt-6-sol-and-luna.

Both models also accept image inputs — relevant if vision-based inspection
is ever proposed; no registry format change needed for that.

Acceptance

• Both entries present in both registries with inline source citations.
• Both IDs covered by the test_metadata_flagship_generation lookup loops
  (per review feedback on feat(providers): add current GPT budgets to OpenAI and Azure registries #539, no entry lands without loop coverage).