Add websockets to talk extras and update dependency groups

[kovtcharov]

Summary

The streaming transcription feature (#372) requires the websockets package for the OpenAI SDK's beta.realtime module. Currently websockets is not listed in any extras group in setup.py.

Changes Required

File	Change	Effort
setup.py	Add websockets to the talk extras group	Low
setup.py	Verify openai version in base deps supports beta.realtime (requires openai >= 1.x)	Low

Current talk extras

"talk": [
    "sounddevice",
    "openai-whisper",
    "kokoro>=0.3.1",
    "soundfile",
    "psutil",
    "pip",
],

Proposed talk extras

"talk": [
    "sounddevice",
    "openai-whisper",
    "kokoro>=0.3.1",
    "soundfile",
    "psutil",
    "pip",
    "websockets",
],

Acceptance Criteria

• websockets added to talk extras
• pip install -e ".[talk]" installs websockets
• OpenAI SDK version in base deps supports beta.realtime