Fix for issue #33

[agent-scarn]

This PR introduces support for selecting a custom AI engine (e.g., self-hosted Llama) alongside OpenAI. Key changes:

1. Created config/ai.php to define the active engine (AI_ENGINE) and credentials/settings for both OpenAI and Llama endpoints.
2. Added App\Services\AiService to abstract chat and audio transcription calls. It delegates to the configured engine: OpenAI via the existing facade, or Llama via HTTP requests.
3. Updated all AI-related Actions (GenerateSubtitles, TranslateSubtitles, GenerateChapters, GenerateSummary) to inject and use AiService instead of directly calling the OpenAI facade.

With these changes, you can now set AI_ENGINE=openai or AI_ENGINE=llama in your .env (along with LLAMA_API_URL and LLAMA_API_KEY) to switch between providers seamlessly.

Closes #33