feat(speech): Integrated a "SpeechEmotionModel" using HuggingFace {superb/wav2vec2-base-superb-er} with config-driven setup and standardized output schema.

[VedaSiddhartha010]

/// Integrated a "SpeechEmotionModel" for Voice/Speech Analysis for emotion recognition.

What’s Implemented : ----

• Introduced a new SpeechEmotionModel implemented as a torch.nn.Module.
• Integrated HuggingFace audio classification pipeline using:
  • superb/wav2vec2-base-superb-er
• Added configuration support in config.yaml for flexible model and device selection.
• Designed the model to be fully config-driven, ensuring easy extensibility and maintainability

Output Format : ---

The model returns structured output consistent with the existing pipeline:

{
  "emotion": {
    "label": "...",
    "score": "..."
  }
}