Adds support for basic turn detection

[Nash0x7E2]

In this PR, we introduce the class TurnDetection, which allows the implementation of various turn detection models. By default, we also include a small example using fal/smart-turn, which can be found in examples/example_turn_detection.py.

Params:

• mini_pause_duration
• max_pause_duration
• is_detecting
• on
• process_audio
• start
• stop
• emit

When using the TurnDetection class, the following events can be emitted and observed:

class TurnEvent(Enum):
    """Events that can occur during turn detection."""

    SPEECH_STARTED = "speech_started"
    SPEECH_ENDED = "speech_ended"
    TURN_STARTED = "turn_started"
    TURN_ENDED = "turn_ended"
    MINI_PAUSE_DETECTED = "mini_pause_detected"
    MAX_PAUSE_REACHED = "max_pause_reached"

The base agents/agent.py class has also been modified to support the turn_detection parameter in __init__.