feat: expand OpenAI API compatibility with extended schemas, streaming usage, and advanced parameters

[raiyanu]

Summary

Makes the server fully OpenAI-compatible so it works seamlessly as a drop-in provider for OpenAI SDKs, LangChain, n8n, NextChat, OpenWebUI, etc.

Key Changes

• Dual Endpoints: Added root paths alongside /v1 (e.g. POST /chat/completions, GET /models, POST /embeddings).
• OpenAI Schema: Supports all standard request parameters (stream_options, response_format, stop, max_tokens, etc.) with extra="allow" to prevent 422 errors.
• Response & Streaming: Compliant chat.completion and chat.completion.chunk formatting (with usage in streaming, system_fingerprint, and standard finish reasons).
• Model Aliases & Thinking: Added aliases (gpt-4o, gpt-3.5-turbo, deepseek-reasoner) and auto-enables DeepThink for reasoning models.
• CORS & Auth: Added CORS middleware and optional API_KEY verification.

[raiyanu]

@sums001