Support [AI-59](https://linear.app/affine-design/issue/AI-59)
Deprecated use of the perplexity model. Makes chat and search use the same prompt content.
<!-- This is an auto-generated comment: release notes by coderabbit.ai -->
- **New Features**
- Added support for two new AI models: Claude Opus 4 and Claude Sonnet 4, enabling enhanced text and image input capabilities.
- Introduced a new chat prompt with detailed instructions for specialized AI interactions within AFFiNE.
- **Refactor**
- Improved prompt management for AI chat and search features by centralizing shared settings for better consistency.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
Close [AI-125](https://linear.app/affine-design/issue/AI-125)
<!-- This is an auto-generated comment: release notes by coderabbit.ai -->
## Summary by CodeRabbit
- **New Features**
- Added new provider configurations `geminiVertex` and `anthropicVertex` for Google Vertex AI in backend schema, provider classes, and admin config.
- Introduced `GeminiVertexProvider` and `AnthropicVertexProvider` classes supporting Vertex AI models with specific capabilities.
- Expanded model options for transcription prompts with newer Gemini models.
- Re-exported provider modules to include Vertex AI variants.
- **Improvements**
- Extended provider architecture to support separate Vertex AI configurations and models.
- Updated test setup to replace deprecated provider references with new Vertex variants.
- Consolidated environment variables for server testing with a single `SERVER_CONFIG`.
- **Bug Fixes**
- Updated mock models and import references in tests to align with new provider classes.
- **Chores**
- Added `@ai-sdk/google-vertex` dependency for Vertex AI support.
- Updated dependency list to include `@ai-sdk/google-vertex`.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->