fix(openai): use codex context limits for gpt-5.6

[nabilfreeman]

Issue for this PR

Closes #36247

Type of change

• Bug fix
• New feature
• Refactor / code improvement
• Documentation

What does this PR do?

OpenCode inherits the direct API's 1.05M limits for GPT-5.6 even when OpenAI authentication routes requests through Codex OAuth. The Codex catalog reports a 372K input budget for Sol, Terra, and Luna.

This extends the existing GPT-5.5 OAuth override to use context: 500000, input: 372000, and output: 128000 for all three GPT-5.6 models. API-key authentication continues using the direct API metadata unchanged.

How did you verify your code works?

• bun test test/plugin/codex.test.ts (17 pass)
• bun typecheck from packages/opencode
• pre-push bun turbo typecheck (30 successful)

Screenshots / recordings

Not applicable; this is a provider metadata fix with no UI changes.

Checklist

• I have tested my changes locally
• I have not included unrelated changes in this PR

[nabilfreeman]

@rekram1-node tagging you because I know you worked on adding 5.6 yesterday to models.dev.

[nabilfreeman]

Correction on the limit mapping: we initially interpreted codex debug models reporting context_window: 372000 as OpenCode's total context value.

Comparing it with GPT-5.5 clarified the convention. Codex reports 272000 for GPT-5.5, while OpenCode's existing OAuth override maps that to input: 272000, then adds the 128000 output allowance for context: 400000.

This PR now follows the same mapping for GPT-5.6:

context: 500000
input:   372000
output:  128000

The correction is in abc5bf44f7. The focused Codex tests and package typecheck pass, and the pre-push workspace typecheck completed all 30 tasks successfully.

[github-actions]

Thanks for updating your PR! It now meets our contributing guidelines. 👍

[timrichardson]

note: context limits for codex plans have been lowered to gpt-5.5 levels: https://x.com/thsottiaux/status/2076495156757577895?s=20 for the next "few days"

[nabilfreeman]

note: context limits for codex plans have been lowered to gpt-5.5 levels: https://x.com/thsottiaux/status/2076495156757577895?s=20 for the next "few days"

🥲🥲🥲🥲🥲🥲🥲🥲🥲🥲🥲

This reinforces the need to remotely configure the codex limits à la anomalyco/models.dev#3183