Add realtime buffer size for voice agent

[rachwalk]

Is your feature request related to a problem? Please describe.
As described in this comment: #393 (comment) the VoiceRecognitionAgent src/rai/rai/agents/voice_agent.py currently uses a set size for sample buffer:

        with self.sample_buffer_lock:
            self.sample_buffer.append(indata)
            if not self.recording_started and len(self.sample_buffer) > 5:
                self.sample_buffer = self.sample_buffer[-5:]

It could be beneficial to be able to configure that size during runtime.

Describe the solution you'd like
A connector for configuration could be added, with the agent receiving messages during runtime and setting appropriate parameter based on what it receives.

Describe alternatives you've considered
All alternatives require some form of a runtime communication with the Agent process, which in turn given the current RAI API would require a connector.

Additional context
N/A