Support 0shape tensor

[wyg1997]

本PR支持 0shape Tensor的创建、计算，例如：

x = flow.ones((4, 0, 4))
y = flow.ones((4, 2, 4))
z = flow.cat([x, y], dim=1)
print(z.shape)  # flow.Size([4, 2, 4])

TODO

• 去掉indexing中的CHECK，支持 0shape 的slice Dev advanced indexing #5559

相关issue

#5399 #5537

[github-actions]

CI failed, removing label automerge

[github-actions]

Speed stats:

GPU Name: GeForce GTX 1080 

PyTorch resnet50 time: 137.5ms (= 6877.0ms / 50, input_shape=[16, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 126.9ms (= 6343.2ms / 50, input_shape=[16, 3, 224, 224], backward is enabled)
Relative speed: 1.08 (= 137.5ms / 126.9ms)

PyTorch resnet50 time: 83.5ms (= 4176.4ms / 50, input_shape=[8, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 76.7ms (= 3836.1ms / 50, input_shape=[8, 3, 224, 224], backward is enabled)
Relative speed: 1.09 (= 83.5ms / 76.7ms)

PyTorch resnet50 time: 57.1ms (= 2854.8ms / 50, input_shape=[4, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 48.8ms (= 2442.3ms / 50, input_shape=[4, 3, 224, 224], backward is enabled)
Relative speed: 1.17 (= 57.1ms / 48.8ms)

PyTorch resnet50 time: 45.7ms (= 2287.0ms / 50, input_shape=[2, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 43.2ms (= 2158.5ms / 50, input_shape=[2, 3, 224, 224], backward is enabled)
Relative speed: 1.06 (= 45.7ms / 43.2ms)

PyTorch resnet50 time: 43.5ms (= 2175.3ms / 50, input_shape=[1, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 41.9ms (= 2097.2ms / 50, input_shape=[1, 3, 224, 224], backward is enabled)
Relative speed: 1.04 (= 43.5ms / 41.9ms)

[hjchen2]

这个不应该删吧，我们就是不支持负数step

[liufengwei0103]

这个不应该删吧，我们就是不支持负数step

已修改