Revert "Clamp GPT-5 max output tokens to 20% of context window"

[mrubens]

Reverts #8495

---

Important

Reverts clamping of GPT-5 model max output tokens to 20% of context window, allowing full configured max tokens instead.

• Behavior:
  • Reverts clamping of GPT-5 model max output tokens to 20% of context window in getModelMaxOutputTokens() in api.ts.
  • GPT-5 models now use their full configured max tokens, bypassing the 20% cap.
• Tests:
  • Update api.spec.ts to reflect the removal of the 20% cap for GPT-5 models.
  • Tests for various GPT-5 model IDs to ensure full max tokens are used.
  • Non-GPT-5 models still tested for 20% cap enforcement.

^{This description was created by for 87700ba. You can customize this summary. It will automatically update as commits are pushed.}

[roomote]

No new issues found. The revert is clean and complete - all tests pass and the changes correctly restore the 20% cap for GPT-5 models.