Product backlog

### Description of feature

## Nimbus inference repo
- [x] Convert weights from tf to torch with [ivy](https://github.com/unifyai/ivy)
- [x] #27

## Nimbus inference class
- [x] Adapt from class from [Nimbus repo](https://github.com/angelolab/Nimbus/blob/main/src/cell_classification/inference.py)
- [x] Add tile & stitch inference
- [x] Add different data source structures (i.e. ome-tiff directories)
- [x] Add inference notebook comparable to [this one](https://github.com/angelolab/Nimbus/blob/main/templates/1_Nimbus_Predict.ipynb)
- [ ] Add easy-to-use and hackable commandline interface
- [x] Change user input half-resolution to the actual magnification level of the images and fix type casting mentioned in #26.

## Viewer Widget
- [x] Add viewer widget from [here](https://github.com/angelolab/Nimbus/blob/main/src/cell_classification/viewer_widget.py)
- [x] Make viewer widget work with less than three channels
- [x] Build a second viewer widget with channels on the left, instance map on the right and all cells gray except the one we're interested in in white.
- [x] Adapt the viewer to show marker images and gold-standard groundtruth next to each other

## General project
- [x] Make the data available (EMBL)
- [x] Clean up plot notebooks
- [x] Update the documentation and the pypi distribution