I have been bitten by this in CI where I ran the script after adding a fallback to fallback_ops.py and then CI fails first on XPU, and then later when I try to merge, it fails on MPS just cuz the autogen only added the fallback to the cpu and cuda headers (as I build with CUDA).

I would like to request that the script updates for all the c_shim_*_.h headers cuz that would be my intention anyway, but I'm curious to know if there's a particular reason why we wouldn't want that.

cc @chauhang @penguinwu @avikchaudhuri @gmagogsfm @zhxchen17 @tugsbayasgalan @angelayi @suo @ydwu4 @desertfire @chenyang78 @yushangdi @benjaminglass1