Add min cut partitioner for AOT+nvFuser (#88204)
Here we mark most of `torch.ops.nvprims` as something that can be recomputed in the backward passes (and hopefully fused).
TODO:
- [x] Add a test after https://github.com/pytorch/pytorch/pull/88186 is merged
Pull Request resolved: https://github.com/pytorch/pytorch/pull/88204
Approved by: https://github.com/jjsjann123, https://github.com/jansel