- Add auto_detect.rs module for parallel provider testing
- Add /config/detect-provider API endpoint
- Add DetectProviderRequest and DetectProviderResponse structs
- Update OpenAPI specifications
- Backend compiles successfully

Part of PR #4950 + #5147 integration

- Add ApiKeyTester component calling /config/detect-provider
- Update ProviderGuard with ApiKeyTester in onboarding
- Add provider icons and regenerate API client
- Show progress and results during detection

Integration of PR #4950 + #5147

- Replace format-based detection with parallel testing approach
- Test all providers simultaneously using tokio::spawn for better accuracy
- Simplify API response (404 for no match instead of structured errors)
- Remove complex Ollama rejection logic - let parallel testing handle it
- Should properly detect Anthropic, OpenAI, Google, Groq, xAI, and Ollama
- Maintains existing Quick Setup UI with improved backend detection

This integrates the robust parallel testing approach from PR #5147
into our Quick Setup onboarding flow for better API key detection.

- Add detect_cloud_provider_from_api_key() function that excludes Ollama
- Add /config/detect-cloud-provider endpoint for Quick Setup use
- Update ApiKeyTester to use cloud-only detection endpoint
- Remove frontend Ollama rejection since backend won't return it
- Ensures Quick Setup only works with cloud API providers
- Maintains full detection (including Ollama) for other use cases

This prevents unwanted Ollama fallbacks in Quick Setup while preserving
the parallel testing approach from PR #5147 for better accuracy.

The pure parallel approach from PR #5147 was testing Anthropic keys against
OpenAI API, causing incorrect rejections. This hybrid approach:

1. Detects key format first (sk-ant- → anthropic)
2. Tests the likely provider first with the key
3. Falls back to parallel testing if format detection fails
4. Includes OpenRouter support (sk-or- format)

This should correctly route sk-ant- keys to Anthropic API for proper validation
while maintaining the robustness of parallel testing for unknown formats.

Benefits:
- Anthropic keys test against Anthropic API (not OpenAI)
- Faster detection (test likely provider first)
- Robust fallback (parallel testing if needed)
- Supports more providers (OpenRouter)

- Enhanced frontend format detection for all major providers (Anthropic, OpenRouter, OpenAI, Google, Groq, xAI)
- Added smart workarounds for backend parallel testing issues
- Prevent unwanted redirects during API key testing by maintaining userInActiveSetup state
- Reject Ollama fallback in Quick Setup to ensure cloud-only provider detection
- Configure providers directly when backend detection fails but format is recognized
- Use correct model names: claude-3-haiku-20240307 for Anthropic, anthropic/claude-3-haiku for OpenRouter
- Add input validation to prevent console log injection issues
- Provide detailed error messages and suggestions for failed key validation
- Integrate parallel testing approach from PR #5147 with frontend filtering

Fixes the core issue where users entering incorrect API keys would be redirected
away from setup screen, now they stay on setup with clear error feedback.