fix: add openai-compatible provider support for token usage display

[roomote]

Summary

This PR fixes the issue where token usage displays as 0 when using OpenAI Compatible API providers.

Problem

When users select "OpenAI Compatible" as their API provider and use a custom API endpoint, the token usage counter shows 0 instead of the actual token count.

Solution

Added proper support for the "openai-compatible" provider throughout the codebase:

• Added "openai-compatible" as a valid provider in the API handler
• Updated type definitions and schemas to include the new provider
• Ensured the provider uses the same OpenAiHandler as the standard "openai" provider
• Added comprehensive test coverage

Changes

• Added "openai-compatible" to provider types and schemas
• Updated buildApiHandler to route "openai-compatible" to OpenAiHandler
• Updated ProfileValidator to handle the new provider
• Updated webview-ui model selection to support the new provider
• Added tests to verify token usage tracking works correctly

Testing

• ✅ All existing tests pass
• ✅ New tests added for openai-compatible provider
• ✅ Type checking passes
• ✅ Linting passes

Fixes #8543

---

Important

Adds support for openai-compatible provider, ensuring correct token usage tracking and behavior similar to openai.

• Behavior:
  • Adds support for openai-compatible provider, ensuring token usage is tracked correctly.
  • Routes openai-compatible to OpenAiHandler in buildApiHandler() in index.ts.
• Schemas and Types:
  • Adds openai-compatible to customProviders and providerNames in provider-settings.ts.
  • Updates providerSettingsSchemaDiscriminated to include openai-compatible.
• Validation:
  • Updates getModelIdFromProfile() in ProfileValidator.ts to handle openai-compatible.
• UI:
  • Updates getSelectedModel() in useSelectedModel.ts to support openai-compatible.
• Testing:
  • Adds openai-compatible.spec.ts to test OpenAiHandler creation and token usage tracking for openai-compatible.

^{This description was created by for c48c870. You can customize this summary. It will automatically update as commits are pushed.}

[roomote]

Conducting self-review like a linter with existential dread—scanning for off-by-one realities.

[roomote]

[P3] Add unit test coverage for the new 'openai-compatible' branch here to prevent regressions. A small test asserting the hook returns the configured openAiModelId and custom model info would lock in the behavior.

[roomote]

[P3] Consider adding an allow-list test case for 'openai-compatible' to validate that model gating (allowAll vs. explicit models) works identically to 'openai'. This will help catch configuration regressions.