Add new autotest

[BBuf]

• oneflow.experimental.reshape
• oneflow.experimental.Tensor.reshape
• oneflow.experimental.squeeze
• oneflow.experimental.Tensor.squeeze
• oneflow.experimental.transpose
• oneflow.experimental.Tensor.transpose
• oneflow.experimental.unsqueeze
• oneflow.experimental.Tensor.unsqueeze
• oneflow.experimental.where (没有Bool类型，无法自动测试)
• oneflow.experimental.Tensor.where (没有Bool类型，无法自动测试)
• oneflow.experimental.Tensor.permute
• oneflow.experimental.nn.Hardswish
• oneflow.experimental.nn.PReLU (PReLU oneflow kernel实现有BUG)
• oneflow.experimental.nn.Hardtanh

[github-actions]

Speed stats:

GPU Name: GeForce GTX 1080 

PyTorch resnet50 time: 137.5ms (= 6876.0ms / 50, input_shape=[16, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 124.0ms (= 6201.8ms / 50, input_shape=[16, 3, 224, 224], backward is enabled)
Relative speed: 1.11 (= 137.5ms / 124.0ms)

PyTorch resnet50 time: 83.1ms (= 4154.6ms / 50, input_shape=[8, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 72.2ms (= 3607.8ms / 50, input_shape=[8, 3, 224, 224], backward is enabled)
Relative speed: 1.15 (= 83.1ms / 72.2ms)

PyTorch resnet50 time: 63.3ms (= 3163.9ms / 50, input_shape=[4, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 51.1ms (= 2552.5ms / 50, input_shape=[4, 3, 224, 224], backward is enabled)
Relative speed: 1.24 (= 63.3ms / 51.1ms)

PyTorch resnet50 time: 46.8ms (= 2339.8ms / 50, input_shape=[2, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 52.5ms (= 2622.9ms / 50, input_shape=[2, 3, 224, 224], backward is enabled)
Relative speed: 0.89 (= 46.8ms / 52.5ms)

PyTorch resnet50 time: 42.8ms (= 2142.2ms / 50, input_shape=[1, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 53.8ms (= 2688.8ms / 50, input_shape=[1, 3, 224, 224], backward is enabled)
Relative speed: 0.80 (= 42.8ms / 53.8ms)