feat: hardcode temperature to 1 for GPT-5 models

[roomote]

This PR hardcodes the temperature to 1 for all GPT-5 model variants as requested by @mrubens in #6819.

Changes

• Modified getModelParams function in src/api/transform/model-params.ts to set temperature to 1 for models starting with "gpt-5"
• Added comprehensive test coverage for GPT-5 temperature hardcoding

Testing

• Added 8 new test cases specifically for GPT-5 temperature behavior
• All existing tests continue to pass
• Verified that temperature is correctly hardcoded to 1 for:
  • gpt-5-2025-08-07
  • gpt-5-mini-2025-08-07
  • gpt-5-nano-2025-08-07

Context

This change was requested in PR #6819 to ensure consistent behavior for GPT-5 models by always using a temperature of 1, regardless of user settings or defaults.

Fixes: Follow-up to #6819

---

Important

Hardcodes temperature to 1 for GPT-5 models in getModelParams and adds tests to verify this behavior.

• Behavior:
  • Hardcodes temperature to 1 for GPT-5 models in getModelParams in model-params.ts.
  • Applies to models starting with "gpt-5" in OpenAI format.
• Testing:
  • Adds 8 test cases in model-params.spec.ts to verify temperature hardcoding for GPT-5 models.
  • Tests ensure temperature is set to 1 for various GPT-5 model variants and not altered for non-GPT-5 models or non-OpenAI formats.

^{This description was created by for d8092e3. You can customize this summary. It will automatically update as commits are pushed.}

[roomote]

Reviewing my own code is like debugging in a mirror - everything looks backwards but the bugs are still mine.

[roomote]

Is the placement after the o1/o3-mini handling intentional? It might be cleaner to group all OpenAI model-specific temperature handling together, perhaps in a separate helper function for better maintainability.

Also, could we add a comment explaining why temperature is hardcoded to 1 for GPT-5 models, similar to lines 143-144 for o1/o3-mini?

[roomote]

Great test coverage! Consider adding a test case for potential future GPT-5 variants (e.g., "gpt-5-turbo") to ensure the startsWith("gpt-5") logic handles all GPT-5 family models correctly.

Also, for consistency, the test descriptions could follow a uniform pattern like "hardcodes temperature to 1 for..." and "preserves user temperature for..."