R27

llama.cpp updated to b5760, hopefully the CUDA 12.8 backend plugin works i've not had time to test it.
All settings now apply without a model reload/restart except a few which will ask if you would like to reload the model.

I forgot to push the changes so the source is incorrect on this release.