Migrate PDF Processing Core from Marker to Docling

Closes #109 ## Changes Made - Migration of PDF processing from marker to Docling in all the parsing endpoints - Rewrote image handling functions in [`omniparse/utils.py`](https://github.com/VyoJ/omniparse/blob/main/omniparse/utils.py) to correctly interface with Docling's `PictureItem` class - Updated `README.md`, documentation and `pyproject.toml` to include Docling in place of Marker - Updated `torch` version to >=2.6.0 to prevent a vulnerability in `torch.load()` and upgraded to a corresponding version of `transformers` that did not give issues in using Florence-2 - Ported dependency management from `poetry` to `uv` for faster setup - Removed redundant or unused code from `__init__.py` files ## Testing - Verified outputs from all the endpoints as working - Made sure the Gradio demo and Colab notebooks work as expected