0.2.0

Features

• Flexible model output shapes: The library now supports tensors with arbitrary dimensions, making it compatible with a wider range of object detection models. The NIF dynamically handles tensor shapes passed from Elixir.
• Configurable run/2 options: The run/2 function now accepts a keyword list of options, including :prob_threshold, :iou_threshold, and :transpose for more flexible usage.
• Improved documentation and tests: Added comprehensive documentation for the new API and included tests for various tensor shapes and the :transpose option.