Dafne

This is a bugfix release that solves a problem in the loading of advanced dicom.
It also incorporates some changes in the dafne_dl package which allow for more flexible image normalization options.

This release fixes some crashes during usage.

Wow, this update really feels like magic! Being able to just improve your ROIs with a single click, it's the dream right?

Well, it's now possible thanks to the integration into Dafne of the [Segment Anything Model](https://github.com/facebookresearch/segment-anything) (SAM)

Minor bug fixes in the display of images

With this release, following shortly the previous one, we brought some feature enhancements. The most notable ones are:

- 3D visualization of the anatomy and the active ROI
- Slice-wise statistics

Precompiled binaries for windows and Mac-ARM (coming soon) are provided. Please use the pip installation for Linux and Mac-Intel.

In this release, we have now dramatically reduced Dafne's memory footprint by compressing the masks on the non-displayed slices in-memory. This allows for typical reductions of more than 10-fold, with negligible lag.

As other improvements, now the deep learning autosegmentation can be applied to a subregion of the image (for example to only segment one limb when it's not centered in the FOV), and a "Mixed" option for the model source was added, to have both remote and local models available at the same time.

As a downside, the compiled Linux and MacOS-Intel versions don't seem to be working properly. For these platforms, dafne should be installed through pip.