refactor: align HuggingFace provider with established pattern

[roomote]

Summary

This PR refactors the HuggingFace provider implementation to match the established pattern used by other providers that fetch models via network calls (e.g., OpenRouter, Glama, Ollama).

Changes

• Moved fetcher to correct location: Moved huggingface-models.ts from src/services/ to src/api/providers/fetchers/huggingface.ts
• Updated fetcher to return ModelInfo: The fetcher now returns Record<string, ModelInfo> instead of raw HuggingFace model data, consistent with other providers
• Integrated with model cache: Added HuggingFace to RouterName type and integrated it with the modelCache.ts system for memory and file caching
• Updated provider to extend RouterProvider: The HuggingFace provider now extends the RouterProvider base class and uses the fetchModel() pattern
• Removed unnecessary wrapper: Deleted src/api/huggingface-models.ts as it's no longer needed
• Updated webview integration: Modified webviewMessageHandler.ts to use the new pattern with getModels() while maintaining backward compatibility

Benefits

1. Consistency: HuggingFace now follows the same pattern as other providers
2. Caching: Model lists are now cached in memory and on disk
3. Maintainability: Easier to understand and modify when all providers follow the same pattern
4. Type safety: Better integration with TypeScript types

Testing

• ✅ All existing tests pass
• ✅ TypeScript compilation successful
• ✅ Linting checks pass
• ✅ Added HuggingFace to RouterModels mock in webview tests

---

Important

Refactor HuggingFace provider to align with other providers, improving consistency, caching, and maintainability.

• Behavior:
  • Refactor HuggingFace provider to align with other providers like OpenRouter and Glama.
  • huggingface-models.ts moved to src/api/providers/fetchers/huggingface.ts.
  • Fetcher now returns Record<string, ModelInfo> instead of raw data.
  • Integrated HuggingFace with modelCache.ts for caching.
  • Updated webviewMessageHandler.ts to use getModels() for HuggingFace.
• Classes & Functions:
  • HuggingFaceHandler now extends RouterProvider and uses fetchModel().
  • Removed src/api/huggingface-models.ts as it's redundant.
• Misc:
  • Added HuggingFace to RouterName type in api.ts.
  • Updated ExtensionMessage.ts and WebviewMessage.ts to reflect changes.
  • Updated HuggingFace.tsx to handle new model fetching logic.

^{This description was created by for 86e05ab. You can customize this summary. It will automatically update as commits are pushed.}

[cte]

@roomote-agent Can we get rid of the requestHuggingFaceModels message and just use requestRouterModels? That seems cleaner.

[roomote]

Hi @cte! I see your suggestion about removing requestHuggingFaceModels and using requestRouterModels instead. That makes sense for consistency. Let me implement this change for you.

[ellipsis-dev]

Consider extracting the conversion logic (lines 75–108) into a separate helper for improved readability and maintainability.

^{This comment was generated because it violated a code review rule: irule_tTqpIuNs8DV0QFGj.}

[roomote]

Done! I've successfully removed requestHuggingFaceModels and updated the code to use requestRouterModels instead.

Here's what I changed:

• Removed requestHuggingFaceModels message type from WebviewMessage.ts
• Removed the requestHuggingFaceModels handler from webviewMessageHandler.ts
• Removed huggingFaceModels from ExtensionMessage.ts
• Updated the HuggingFace.tsx component to use requestRouterModels and extract HuggingFace models from the routerModels response

This simplifies the codebase by using a single unified message type for all router-based model providers. All tests are passing and CI checks have completed successfully.

[daniel-lxs]

Done by #6228