updated to GPT 4o mini

[GioPalusa]

I've updated to GPT 4o mini, which has more tokens and is way more cost efficient and faster than earlier models.

• Added API for translating text into SRT files using OpenAI's GPT-4o-mini model.
• Implemented token-based grouping for efficient API requests (groupSegmentsByTokenLength).
• Added streaming response parsing with parseStreamedResponse to handle large translations.
• Introduced retries and error handling for OpenAI API rate limits.
• Updated logging for better debugging and chunk handling.
• Updated dependencies, including tiktoken for token counting.