feat(api): add provider-agnostic project AI chat endpoint with provider registry

[flooryyyy]

Motivation

• Provide a single project-scoped AI chat endpoint that can work with multiple AI providers and provider configs.
• Allow selecting a specific provider by slug or falling back to the configured default to simplify client integration.
• Decouple provider-specific request/response handling from route logic via a registry to make adding new providers easier.
• Normalize completions into a stable response shape so callers don't need to handle provider-specific payloads.

Description

• Add POST /api/projects/[id]/ai/chat with request validation and provider selection by explicit providerSlug or default provider, returning a normalized completion payload (src/app/api/projects/[id]/ai/chat/route.ts).
• Introduce shared AI types and normalized response contract including AiProviderConfig, ChatCompletionRequest, and NormalizedChatCompletion (src/app/lib/ai/providers/types.ts).
• Add a provider registry that resolves provider configs and dispatches chat completion to the correct executor by provider type (src/app/lib/ai/registry.ts).
• Implement OpenAI/OpenAI-compatible and Anthropic executors that handle provider-specific APIs, extract model/usage/content, and return the normalized shape (src/app/lib/ai/providers/openai.ts, src/app/lib/ai/providers/anthropic.ts).

Testing

• Ran static checks with npm run check which succeeded.
• Ran automated tests with npm run test which passed (55 tests passed).

---

Codex Task