Enable coarse to fine training

[ichsan2895]

Hello, Since this repo is using method from splatfacto, recently nerfstudio's developer change 2 hyperparameters for activating coarse-to-fine training. See this thread for disscussion : Enable coarse-to-fine training

It proved increase the quality a little bit in original splatfacto. Jb-ye and me already proved in some datasets. But, I never tested in this repo. I believe it also works in this repo too.

[pierotofy]

Thanks @ichsan2895 ! If it worked in splatfacto, at this point it should work in opensplat also.