Google Dialog / Gemini Live API open issues.

[pragmatrix]

• Billing seems to be currently missing the distinction between text and audio input tokens. If we want a disjoint computation of cached text / audio input tokens, the library gemini-live needs to be extended to support that.
• Test prompting. Prompt tests #70
• Test billing.
• Test further features:
  • Input Transcription
  • Output Transcription
• Upstream PR(s) for the changes in https://github.com/pragmatrix/gemini-live-rs
• Datacenter selection / GDPR / DSGVO compliance
• Review ServiceEvents parity with openai-dialog
• Are more ServiceEvents events needed? Turn complete events #79