Torch-scatter

What's Changed
* Fix NaN treatment in `logsumexp` (https://github.com/rusty1s/pytorch_scatter/pull/369)
* Optimize `segment_coo` and `segment_csr` for `torch.bfloat16` and `torch.half` (https://github.com/rusty1s/pytorch_scatter/pull/375)
* Add PyTorch 2.1.0 support (https://github.com/rusty1s/pytorch_scatter/pull/395)

New Contributors
* jamesmyatt made their first contribution in https://github.com/rusty1s/pytorch_scatter/pull/376
* AgarwalSaurav made their first contribution in https://github.com/rusty1s/pytorch_scatter/pull/377

**Full Changelog**: https://github.com/rusty1s/pytorch_scatter/compare/2.1.1...2.1.2

* Added support for PyTorch 2.0

* Improved runtime of `scatter_min`, `scatter_max` and `scatter_mul` CUDA kernels (336)
* Enabled ROCm build (325)
* Added bf16 support (316)

* PyTorch 1.10 support
* Allow loading of CPU wheels in a PyTorch CUDA installation https://github.com/rusty1s/pytorch_scatter/issues/229 (thanks to aphedges)
* Removed unnecessary `eps` argument in `scatter_softmax` https://github.com/rusty1s/pytorch_scatter/issues/238 (thanks to Veason-silverbullet)
* Resolved `torch.floor_divide` deprecation warnings https://github.com/rusty1s/pytorch_scatter/issues/236 (thanks to digital-idiot)
* Added a `dim_size` argument to `scatter_softmax` and `scatter_log_softmax` https://github.com/rusty1s/pytorch_scatter/pull/243 (thanks to mallamanis)

* Half-precision support (`torch.half`) for all operators in `torch-scatter`

* Prevent division by zero in `scatter_std`
* Fixed a bug in which an "index out of bounds" error occured in case of `dim_size=None`
* Fixed a bug in which absolute paths were added to `SOURCES.txt`