feat: add GPT-5.4 Codex Spark fast models

[roomote-v0]

Related GitHub Issue

Closes: #12112

Description

This PR attempts to address Issue #12112 by adding GPT-5.4 fast/spark model variants to the OpenAI Codex provider. Feedback and guidance are welcome.

Following the existing gpt-5.3-codex-spark pattern, this adds two new models:

• gpt-5.4-codex-spark - Fast, text-only coding model based on GPT-5.4
• gpt-5.4-mini-codex-spark - Fast, text-only coding model based on GPT-5.4 Mini

These spark variants have the same characteristics as the existing 5.3 spark model:

• Smaller context window (128k vs 1M/400k)
• Smaller max output tokens (8,192 vs 128,000)
• Text-only (no image support)
• Reasoning effort levels: low, medium, high, xhigh
• Subscription-based pricing (no per-token costs)

Test Procedure

• Added new test cases to openai-codex.spec.ts verifying model IDs resolve correctly
• Added specific tests validating spark-specific limits (128k context, 8192 max tokens, no images)
• All 13 tests pass: cd src && npx vitest run api/providers/__tests__/openai-codex.spec.ts

Pre-Submission Checklist

• Issue Linked: This PR is linked to an approved GitHub Issue
• Scope: Changes are focused on the linked issue
• Self-Review: Performed a self-review of the code
• Testing: New tests added for the spark model variants
• Documentation Impact: No documentation updates required
• Contribution Guidelines: I have read the Contributor Guidelines

Documentation Updates

• No documentation updates are required.

Additional Notes

The spark model specs (context window, max tokens, capabilities) are based on the existing gpt-5.3-codex-spark pattern. If OpenAI publishes different specs for GPT-5.4 spark variants, the values may need adjustment.

Interactively review PR in Roo Code Cloud