fix: replace axios with fetch to respect VSCode proxy settings

[roomote]

Description

This PR fixes an issue where the OpenAI Compatible provider's model fetching doesn't respect VSCode's proxy settings (like PAC files), while the chat completions API calls do.

Problem

• The getOpenAiModels function uses axios.get to fetch available models
• The OpenAI SDK (used for chat completions) uses the native fetch API internally
• axios doesn't automatically respect VSCode's proxy settings, but fetch does
• This causes model discovery to fail when behind a proxy, even though chat completions work

Solution

• Replace axios.get with fetch in getOpenAiModels function
• Also updated getLmStudioModels for consistency
• Updated tests to mock fetch instead of axios

Testing

• All existing tests pass
• The change ensures both model fetching and chat completions use the same HTTP mechanism

Fixes #6991

---

Important

Replaced axios with fetch in getOpenAiModels and getLmStudioModels to respect VSCode's proxy settings, updating tests accordingly.

• Behavior:
  • Replace axios.get with fetch in getOpenAiModels and getLmStudioModels to respect VSCode's proxy settings.
  • Ensures model fetching and chat completions use the same HTTP mechanism.
• Testing:
  • Update tests in openai.spec.ts to mock fetch instead of axios.
  • Tests verify correct handling of URLs, headers, and error scenarios.
• Misc:
  • Remove axios import from openai.ts and lm-studio.ts.

^{This description was created by for 61d3f94. 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]

Consider adding error logging here to help debug issues when model fetching fails. Currently we silently return an empty array for non-ok responses, which might make troubleshooting difficult:

Suggested change

	return []
	if (!response.ok) {
	console.error(`Failed to fetch OpenAI models: ${response.status} ${response.statusText}`);
	return []
	}

[roomote]

For consistency with the LM Studio implementation, should we include a Content-Type header here? While it's not strictly necessary for GET requests, maintaining consistency across our fetch implementations would be good:

Suggested change

	headers: headers,
	const response = await fetch(`${trimmedBaseUrl}/models`, {
	method: "GET",
	headers: {
	"Content-Type": "application/json",
	...headers
	},
	})

[roomote]

Would it be helpful to add a reference to the issue this fixes for future maintainability?

Suggested change

	// This matches how the OpenAI SDK makes requests internally
	// Use fetch instead of axios to respect VSCode's proxy settings
	// This matches how the OpenAI SDK makes requests internally
	// Fixes #6991: Model fetching now respects VSCode proxy settings (PAC files, etc.)