Suggestion: Use Git LFS, and/or allow choosing where to store the model package

**Is your feature request related to a problem? Please describe.**
Model files (.zip) can grow big, especially trees can reach multiple gigabytes (e.g. https://github.com/dotnet/machinelearning-modelbuilder/issues/1535#issuecomment-900829228 has 6 gigs). Git does not like big files. Storing locally on dev machine is an alternative but developer machines might not have back up solutions as we tend to rely on Git. Network drives (NAS) are better because they typically have backup solutions running. Developers machine might also run out of disk space. The issue is more important if model versioning is added to model builder.

**Describe the solution you'd like**
a) Create/update .gitattributes when generating models to save them in Large File Storage
b) allow selecting location of where .zip file will be stored, such as a shared network disk. In a more complex version it could also be an online location (uploading the zip files might require fairly complex configuration)

**Describe alternatives you've considered**
Alternative for B: Allow modifying .consumption template so developers can update the shared network disk or the online location.

Re-training pipelines may also have some solutions related to this. If so, then it would be nice to see a general roadmap.

**Additional**
Updating AI models may be necessary and it would be nice to be able to do it without updating the client software. Or, large AI files might be downloaded only on demand.