Efficientv2-unet

This release allows using the [efficientv2-unet](https://github.com/DBM-MCF/efficientv2-unet) v0.0.4, which can be installed via pip.

Some modifications:

- New training batch size: reduce that number if you have a GPU with little memory
- Make sure you also update to the latest [efficientv2-unet code](https://github.com/DBM-MCF/efficientv2-unet)
- Running prediction via script allows prediction in rectangle ROI
- QuPath downscales the images sent for prediction
- When re-running a training, you can choose to reset the training data automatically

First release!
Don't forget to also get the latest [QuPath extension for cellpose](https://github.com/BIOP/qupath-extension-cellpose).

**Full Changelog**: https://github.com/DBM-MCF/efficientv2-unet/compare/v0.0.4...v0.0.5

**Full Changelog**: https://github.com/DBM-MCF/efficientv2-unet/compare/v0.0.3...v0.0.4

**Full Changelog**: https://github.com/DBM-MCF/efficientv2-unet/compare/v0.0.2...v0.0.3