move logical op dostring, and refine docstr file

[doombeaker]

这个 #8064 增加 docstr 文件时，忘记在 __init__.py 中导入新的文件了。导致 docstr 没有生效：https://oneflow.readthedocs.io/en/master/oneflow.html#oneflow.logical_xor

本 PR 在 __init__.py 里添加了导入。顺便修改了一些细节（给 oneflow._C.xxx 添加 docstr 是以前没有讲究的做法，后来看到其它PR要求，直接给面向用户的 API 添加 docstring），把另外文件的 logical op 集中到了一起。

[github-actions]

Speed stats:

GPU Name: NVIDIA GeForce GTX 1080 

❌ OneFlow resnet50 time: 130.4ms (= 13044.9ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 144.2ms (= 14416.0ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.11 (= 144.2ms / 130.4ms)

OneFlow resnet50 time: 78.6ms (= 7860.9ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 91.0ms (= 9100.4ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.16 (= 91.0ms / 78.6ms)

OneFlow resnet50 time: 55.0ms (= 10996.7ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 58.4ms (= 11672.7ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.06 (= 58.4ms / 55.0ms)

OneFlow resnet50 time: 40.3ms (= 8061.5ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 43.6ms (= 8712.6ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.08 (= 43.6ms / 40.3ms)

OneFlow resnet50 time: 34.6ms (= 6914.8ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 36.7ms (= 7342.9ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.06 (= 36.7ms / 34.6ms)

OneFlow swin dataloader time: 0.246s (= 49.168s / 200, num_workers=1)
PyTorch swin dataloader time: 0.151s (= 30.276s / 200, num_workers=1)
Relative speed: 0.616 (= 0.151s / 0.246s)

OneFlow swin dataloader time: 0.066s (= 13.129s / 200, num_workers=4)
PyTorch swin dataloader time: 0.041s (= 8.246s / 200, num_workers=4)
Relative speed: 0.628 (= 0.041s / 0.066s)

OneFlow swin dataloader time: 0.037s (= 7.441s / 200, num_workers=8)
PyTorch swin dataloader time: 0.022s (= 4.394s / 200, num_workers=8)
Relative speed: 0.590 (= 0.022s / 0.037s)

❌ OneFlow resnet50 time: 147.3ms (= 14729.6ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 174.5ms (= 17447.6ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.18 (= 174.5ms / 147.3ms)

OneFlow resnet50 time: 97.5ms (= 9752.6ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 113.7ms (= 11365.7ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.17 (= 113.7ms / 97.5ms)

OneFlow resnet50 time: 71.6ms (= 14320.3ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 88.1ms (= 17615.8ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.23 (= 88.1ms / 71.6ms)

OneFlow resnet50 time: 59.1ms (= 11828.2ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 73.7ms (= 14748.4ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.25 (= 73.7ms / 59.1ms)

OneFlow resnet50 time: 54.5ms (= 10899.9ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 76.2ms (= 15233.3ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.40 (= 76.2ms / 54.5ms)

[github-actions]

CI failed when running job: cuda-benchmark. PR label automerge has been removed

[github-actions]

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/8297/

[github-actions]

Speed stats:

GPU Name: NVIDIA GeForce GTX 1080 

❌ OneFlow resnet50 time: 130.5ms (= 13046.5ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 143.8ms (= 14381.9ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.10 (= 143.8ms / 130.5ms)

OneFlow resnet50 time: 76.9ms (= 7687.9ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 89.0ms (= 8899.2ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.16 (= 89.0ms / 76.9ms)

OneFlow resnet50 time: 54.8ms (= 10963.0ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 62.6ms (= 12515.2ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.14 (= 62.6ms / 54.8ms)

OneFlow resnet50 time: 41.5ms (= 8292.4ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 44.6ms (= 8927.9ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.08 (= 44.6ms / 41.5ms)

OneFlow resnet50 time: 36.9ms (= 7371.5ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 40.6ms (= 8121.8ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.10 (= 40.6ms / 36.9ms)

OneFlow swin dataloader time: 0.245s (= 48.965s / 200, num_workers=1)
PyTorch swin dataloader time: 0.152s (= 30.416s / 200, num_workers=1)
Relative speed: 0.621 (= 0.152s / 0.245s)

OneFlow swin dataloader time: 0.104s (= 20.821s / 200, num_workers=4)
PyTorch swin dataloader time: 0.041s (= 8.260s / 200, num_workers=4)
Relative speed: 0.397 (= 0.041s / 0.104s)

OneFlow swin dataloader time: 0.035s (= 7.076s / 200, num_workers=8)
PyTorch swin dataloader time: 0.024s (= 4.700s / 200, num_workers=8)
Relative speed: 0.664 (= 0.024s / 0.035s)

❌ OneFlow resnet50 time: 146.2ms (= 14615.9ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 173.6ms (= 17356.1ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.19 (= 173.6ms / 146.2ms)

OneFlow resnet50 time: 97.2ms (= 9723.1ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 113.9ms (= 11390.2ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.17 (= 113.9ms / 97.2ms)

OneFlow resnet50 time: 73.4ms (= 14685.9ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 88.1ms (= 17616.2ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.20 (= 88.1ms / 73.4ms)

OneFlow resnet50 time: 60.9ms (= 12188.0ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 74.7ms (= 14930.8ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.23 (= 74.7ms / 60.9ms)

OneFlow resnet50 time: 56.3ms (= 11257.5ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 70.7ms (= 14141.7ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.26 (= 70.7ms / 56.3ms)