fix: handle zero timeout correctly for OpenAI-compatible providers

[roomote]

Summary

This PR fixes the timeout issue with LM Studio and other OpenAI-compatible providers where setting apiRequestTimeout to 0 (intended to disable timeout) was causing an immediate timeout instead.

Problem

The OpenAI SDK interprets a timeout value of 0 as an immediate timeout rather than "no timeout". When users set apiRequestTimeout to 0 expecting to disable the timeout, requests would fail immediately.

Solution

• Modified the timeout handling in LM Studio, OpenAI, and Ollama providers to pass undefined instead of 0 to the OpenAI client when the timeout is set to 0
• The OpenAI SDK correctly interprets undefined as "no timeout"
• Updated all related tests to verify the correct behavior

Changes

• src/api/providers/lm-studio.ts: Convert 0 timeout to undefined
• src/api/providers/openai.ts: Convert 0 timeout to undefined
• src/api/providers/ollama.ts: Convert 0 timeout to undefined
• src/api/providers/tests/*.spec.ts: Updated tests to expect undefined for zero timeout

Testing

• ✅ All existing tests pass
• ✅ Added/updated tests specifically for the zero timeout case
• ✅ Linting and type checking pass

Fixes #7366

---

Important

Fixes timeout handling by converting 0 to undefined for OpenAI-compatible providers, ensuring no timeout is set.

• Behavior:
  • Fixes timeout handling for LmStudioHandler, OpenAiHandler, and OllamaHandler by converting 0 timeout to undefined.
  • OpenAI SDK now correctly interprets undefined as no timeout.
• Files:
  • lm-studio.ts, openai.ts, ollama.ts: Convert 0 timeout to undefined.
  • lm-studio-timeout.spec.ts, ollama-timeout.spec.ts, openai-timeout.spec.ts: Update tests to expect undefined for zero timeout.
• Testing:
  • All existing tests pass.
  • Added/updated tests for zero timeout case.
  • Linting and type checking pass.

^{This description was created by for 3f98467. 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 this pattern intentional? I notice that here in LM Studio and Ollama, we're doing the conversion inline, but in the OpenAI provider, we're using an intermediate variable clientTimeout. Would it be worth standardizing on one approach across all three providers for consistency?

[roomote]

Nice use of an intermediate variable here! Though I'm wondering if we should consider extracting this conversion logic (0 to undefined) into a shared utility function since it's repeated in all three providers? Something like:

Suggested change

	const clientTimeout = timeout === 0 ? undefined : timeout
	const clientTimeout = normalizeOpenAITimeout(timeout)

Where normalizeOpenAITimeout could live in the utils folder and handle this conversion consistently.

[roomote]

Same pattern as LM Studio - inline conversion. Could we consider using the same approach as the OpenAI provider with an intermediate variable for better readability across all providers?