[BugFix][easy] Fix flaky test test_gpt_oss_multi_turn_chat

[lacora]

Purpose

Current test on test_gpt_oss_multi_turn_chat is pretty flaky and fails >50% times.

Example of failed test asking for if to use Celsius or Fahrenheit :
root:test_serving_chat.py:187 lacora: response ChatCompletion(id='chatcmpl-9596f8584920486b8e0e22468e55e606', choices=[Choice(finish_reason='stop', index=0, logprobs=None, message=ChatCompletionMessage(content='Sure! Would you like the temperature in Celsius or Fahrenheit?', refusal=None, role='assistant', annotations=None, audio=None, function_call=None, tool_calls=[], reasoning_content='We need to call the function get_current_weather with city "Dallas", state "TX", unit maybe default? The user didn't specify unit. We can ask for unit? But we can choose default. Probably ask for unit? The user didn't specify. We can ask: "Would you like Celsius or Fahrenheit?" But we can also default to Fahrenheit for US. Let's ask.'), stop_reason=None, token_ids=None)], created=1757467045, model='openai/gpt-oss-20b', object='chat.completion', service_tier=None, system_fingerprint=None, usage=CompletionUsage(completion_tokens=97, prompt_tokens=153, total_tokens=250, completion_tokens_details=None, prompt_tokens_details=None), prompt_logprobs=None, prompt_token_ids=None, kv_transfer_params=None)

Example of successful test outputing function call:
root:test_serving_chat.py:187 lacora: response ChatCompletion(id='chatcmpl-04de7729d5a1444fbc7072b029b4a945', choices=[Choice(finish_reason='tool_calls', index=0, logprobs=None, message=ChatCompletionMessage(content='{"city":"Dallas","state":"TX","unit":"celsius"}', refusal=None, role='assistant', annotations=None, audio=None, function_call=None, tool_calls=[ChatCompletionMessageFunctionToolCall(id='chatcmpl-tool-3190042ee74c466db8b019aa3431dfc0', function=Function(arguments='{"city":"Dallas","state":"TX","unit":"celsius"}', name='get_current_weather'), type='function')], reasoning_content='We need to call the function get_current_weather with city "Dallas", state "TX", unit "celsius".'), stop_reason=200012, token_ids=None)], created=1757466163, model='openai/gpt-oss-20b', object='chat.completion', service_tier=None, system_fingerprint=None, usage=CompletionUsage(completion_tokens=57, prompt_tokens=157, total_tokens=214, completion_tokens_details=None, prompt_tokens_details=None), prompt_logprobs=None, prompt_token_ids=None, kv_transfer_params=None)

Test Plan & Test Result

Fix and run multiple times all succeeded.

---

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.
• (Optional) Release notes update. If your change is user facing, please update the release notes draft in the Google Doc.

[yeqcharlotte]

thanks! cc: @heheda12345