sggao/celllens ? reverse-engineered prompt

Build me a Python package and notebook based workflow for CellLENS, a tool that takes single cell spatial omics data and finds meaningful cell populations by combining gene or protein expression, cell locations, and optional tissue images.

I want it to be easy to install with pip, import as celllens, and run from a few notebooks with clear examples for a basic mode and a full mode that includes images. Please make the main flow feel simple, something like load data, initialize the model, prepare images if available, train it, and get a final embedding or clustering result that I can use to explore cells.

Use GPU support if needed, and make the code friendly for someone following along in Jupyter or Colab. Include a small example dataset or demo path if possible, and make sure the tutorial notebooks actually show how to use the library end to end. If you need to check current package or notebook docs online, go ahead and do that.