Added torch model

[JLrumberger]

What is the purpose of this PR?
Purpose is to provide an inference script/notebook that works on all platforms (linux, windows, macOS).

How did you implement your changes
This PR adds the following:

• the tf2torch port of the nimbus model checkpoint
• class Nimbus which checks the input, loads the model, calculates the normalization dict and does whole-image or shift equivariant tile-and-stitch inference
• the Unet class that implements a residual unet
• the ViewerWidget class that let's the user inspect the predictions.
• the updated 1_Nimbus_Predict.ipynb notebook for low-code inference.

Remaining issues

• ome.tiff inference is still missing and will be added in the coming weeks.
• coverage action fails, Sri will look into this.

[JLrumberger]

This issue explains some of the weird behavior of the github checkout action when trying to use LFS with it.

[JLrumberger]

@srivarra I have a hard time debugging the coveralls action. Could you look into this? It seems that .lcov files are found and uploaded but no report is generated. Thus it could be that the .lcov files are empty or that something else is preventing coveralls from returning a proper report.. Locally I generated .lcov files with pytest and transformed them to .html with coverage html and everything looks fine..

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB