Add support for AAC audio encoding

[JonnyTran]

Fixes #295

Description

This PR adds support for AAC audio format conversion using FFmpeg through the ffmpeg-python binding. This is particularly useful for iOS applications that require AAC format when using OpenAI TTS backend.

Changes

• Added ffmpeg-python as a new dependency
• Modified convert_audio_format to handle AAC conversion
• Moved "aac" from UNSUPORTED_RESPONSE_FORMATS to SUPPORTED_RESPONSE_FORMATS

Dependencies

• Requires ffmpeg-python: pip install ffmpeg-python
• System FFmpeg must be installed (already a requirement)

Testing

• Tested AAC conversion with various input formats
• Verified output plays correctly on iOS Speech Central app
• All existing format conversions continue to work as before

Note

The implementation uses ffmpeg-python for cleaner and safer FFmpeg integration rather than direct subprocess calls. The conversion is done entirely in memory without touching the filesystem.