refactor(models): standardize outputs to structured dictionary format and added config validation.

[VedaSiddhartha010]

This PR standardizes model outputs across the repository by replacing tuple-based returns with structured dictionaries.

Changes : ---

• Updated Whisper Transcript to return {"text", "chunks"} instead of tuple
• Updated Bertweet sentiment model to return {"label", "score", "probabilities"}
• Refactored dependent service layers to handle dict-based outputs
• Added config validation checks to prevent runtime errors

Benefits: ---

• Improved readability and maintainability
• Consistent API design across models
• Easier integration with downstream services and pipelines

Notes : ---

• No functional changes to model predictions
• Only output structure has been modified

[VedaSiddhartha010]

@BasmaElhoseny01 Ma'am can you review this