Add EdgeCrafter Model Support to the Lightning Engine

[kprokofi]

Description
Integrate the EdgeCrafter architecture into the existing Lightning-based training engine to expand the portfolio of supported state-of-the-art edge-optimized models.

Goals

• Add EdgeCrafter model implementation to the Lightning engine
• Integrate with the existing training, validation, and inference pipelines
• Support configuration through the current model registry/configuration system
• Enable checkpoint loading and model export where applicable
• Ensure compatibility with the existing dataset preprocessing and augmentation pipeline

Acceptance Criteria

• EdgeCrafter is available as a supported model in the Lightning engine
• Users can train, validate, and run inference using the existing API
• Configuration follows the existing Lightning engine conventions
• Documentation includes usage instructions and an example configuration
• Unit and integration tests cover the new model