Update on "[inductor] Add ir.Scan and lower aten.cumsum on CUDA"


This adds the `ir.Scan` node (currently only supported on CUDA) which re-uses the existing reduction kernel machinery to support different kinds of non-pointwise ops. Just like reductions it supports prologue and epilogue fusions and has both persistent and non-persistent kernel generation.

Currently this doesn't support the equivalent of `Reduction.create_multilayer` and will instead fall back to eager in those cases. This is because splitting into multiple kernel invocations ends up being far slower than cub's single kernel strategy which matches the performance of a copy kernel.

Fixes pytorch#93631


cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng Xia-Weiwen wenzh
8000
e-nrv jiayisunx ipiszy ngimel yf225 chenyang78 kadeng muchulee8 aakhundov

[ghstack-poisoned]

[BE]: Update ruff to v0.1.7

[DTensor][BE] fix bug in OpStrategy for Tuple output

[ghstack-poisoned]

Update gloo submodule

Update on "Fix Conv Binary Inplace Fusion issue"


**Summary**
Take this Pattern as example
```
  #      ReLU
  #     /   \
  #  Conv1
  #   /      \
  # Conv2
  #   \      /
  #      Add
```
Current `ConvBinaryInplace` check will fail to do the Inplace fusion (use the outplace fusion instead) due to `ReLU` has 2 users. However, if all the users of `ReLU` are ancestor_node of `Conv2`, we should be able to do the `ConvBinaryInplace` fusion. Relax the `ConvBinaryInplace` check in this Diff.

**TestPlan**
```
python -m pytest test_mkldnn_pattern_matcher.py -k test_conv2d_binary_inplace_fusion_pass_cpu
python -m pytest test_mkldnn_pattern_matcher.py -k test_conv2d_binary_inplace_fusion_failed_cpu
```

cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx peterbell10 ipiszy yf225 chenyang78 kadeng muchulee8 aakhundov ColinPeppler

[ghstack-poisoned]

Add _reset_nccl_collective_timeout

[ghstack-poisoned]

Update on "[dynamo] Improve graph break message for copy.deepcopy"


I was curious what hf_T5_generate was trying to deepcopy, so I updated the errror message:
Before:
```
STATS graph_break
  ("'skip function deepcopy in file /home/jansel/conda/envs/pytorch/lib/python3.10/copy.py'', skipped according skipfiles.SKIP_DIRS'", 3)
  ...
```
After:
```
STATS graph_break
  ('copy.deepcopy UserDefinedObjectVariable(GenerationConfig)', 3)
  ...
```

Related issue: pytorch#115122

cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 aakhundov kadeng

[ghstack-poisoned]

Update on "[BE][HigherOrderOp] extract redundant code that unflattens…

… the output"


We need this function to unflatten the variable tracker for HOPs that want pytree output support.

cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 aakhundov kadeng

[ghstack-poisoned]

[aoti] Link to export docs in AOTI docs

Merge branch 'main' into hta-docs