fix: improve OpenRouter model validation to support newer models

[roomote]

This PR attempts to address Issue #9597 where newer OpenRouter models (like gpt-5.1, gemini-3-pro-preview, and grok-4.1-fast:free) were being incorrectly marked as invalid after the model cache was populated.

Problem

The validation logic was too strict - it would reject any model not in the cached list when the cache had more than 1 model. This caused newer models to fail validation after a few minutes when the full model list was fetched.

Solution

1. Increased validation threshold: Changed from >1 to >10 models before validation kicks in, preventing false negatives during initial cache population
2. Added pattern matching: For OpenRouter specifically, added regex patterns to allow known model formats even when not in cache:
   • OpenAI models (e.g., openai/gpt-5.1)
   • Google models (e.g., google/gemini-3-pro-preview)
   • X-AI models (e.g., x-ai/grok-4.1-fast:free)
   • Anthropic, Meta Llama, and Mistral models

Testing

• Added comprehensive tests for the new validation logic
• Tests cover both the new pattern matching and the threshold change
• All existing tests pass

Fixes #9597

Feedback and guidance are welcome!

---

Important

Improve OpenRouter model validation by increasing threshold and adding pattern matching for newer models.

• Behavior:
  • In validate.ts, increase model validation threshold from >1 to >10 models to prevent false negatives.
  • Add regex pattern matching for OpenRouter models in validateDynamicProviderModelId() to allow known formats (e.g., openai/gpt-5.1, google/gemini-3-pro-preview).
• Testing:
  • Add tests in validate.spec.ts for new validation logic, covering pattern matching and threshold changes.
  • Rename validate.test.ts to validate.spec.ts for consistency.
• Misc:
  • Update error messages in validate.ts to use settings:validation namespace.

^{This description was created by for e3c1bc0. You can customize this summary. It will automatically update as commits are pushed.}

[roomote]

Rooviewer   See task on Roo Cloud

Review completed. No issues found in the changes made by this PR.

The implementation looks good:

• The threshold increase from >1 to >10 models is reasonable to avoid false negatives during initial cache population
• The pattern matching for known OpenRouter model formats (OpenAI, Google, X-AI, Anthropic, Meta Llama, Mistral) provides a good fallback for newer models
• The comprehensive test coverage validates both the threshold change and pattern matching behavior
• The file rename from validate.test.ts to validate.spec.ts aligns with project conventions

_{Mention @roomote in a comment to request specific changes to this pull request or fix all unresolved issues.}

[ellipsis-dev]

The regex for x-ai/grok only matches digits (e.g. /^x-ai\/grok-\d+/). To support versions with decimals (e.g. 4.1), consider updating it to /^x-ai\/grok-\d+(\.\d+)?/ for consistency with the OpenAI pattern.

Suggested change

	/^x-ai\/grok-\d+/, // Matches grok-4, grok-4.1, etc.
	/^x-ai\/grok-\d+(\.\d+)?/, // Matches grok-4, grok-4.1, etc.