Refactor audio helper, improve CLI/GUI error handling, and add OpenAI SDK requirement

[SPark1264]

Motivation

• Resolve PyCharm typing/structure issues by simplifying where audio capture lives and keeping the typed OpenAI chat call.
• Ensure microphone capture uses ALSA-compatible tooling on Raspberry Pi by relying on arecord with a clear failure message if not installed.
• Surface runtime errors to users and avoid silent failures in both CLI and GUI flows.

Description

• Moved record_audio_wav(...) to a module-level function and updated BMOAssistant.transcribe_microphone(...) to call it directly.
• Use arecord (ALSA) for recording a mono 16kHz WAV and raise a helpful error if the binary is missing.
• Add invalid-menu-choice handling in run_cli(...) so non-1/2 inputs print a guidance message.
• Add exception handling in GUI worker threads to append Error: <exc> lines to the UI output when background ask_chatbot/transcribe_microphone calls fail.
• Preserve the typed chat call using self.client.chat.completions.create(...) with ChatCompletionSystemMessageParam/ChatCompletionUserMessageParam.
• Add requirements.txt requiring openai>=1.40.0 to match the SDK usage.

Testing

• Ran python -m compileall main.py and the file compiled successfully.

---

Codex Task