Dev reconstruct pad ops

[Flowingsun007]

因为constant_pad1/2/3d op不支持在batch，channel维进行pad，而系统原有pad op同时支持1/2/3D pad且支持在N、C维pad，故此pr使用pad op替换了constant_pad1/2/3d op；

• 使用原有的pad op替换constant_pad1/2/3d op
• pad op的前向反向functor实现
• pad op gradien funcs
• 重构ConstantPad1/2/3d、ZeroPad2d module
• 重构constant pad 1/2/3d auto testcase

重构后，在constant mode下，api对齐torch：https://pytorch.org/docs/stable/generated/torch.nn.functional.pad.html?highlight=pad#torch.nn.functional.pad

[hjchen2]

那之前你写的constant_pad还需要么，是不是可以删了

[Flowingsun007]

那之前你写的constant_pad还需要么，是不是可以删了

嗯，是可以删了
(暂时先不删吧，我看single_client下也用到了这些op，后面在弄一个pr一起删吧）

[github-actions]

Speed stats:

GPU Name: GeForce GTX 1080 

OneFlow resnet50 time: 128.5ms (= 6426.6ms / 50, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 141.8ms (= 7089.1ms / 50, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.10 (= 141.8ms / 128.5ms)

OneFlow resnet50 time: 74.8ms (= 3738.3ms / 50, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 84.9ms (= 4246.6ms / 50, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.14 (= 84.9ms / 74.8ms)

OneFlow resnet50 time: 48.1ms (= 2404.2ms / 50, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 59.1ms (= 2956.9ms / 50, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.23 (= 59.1ms / 48.1ms)

OneFlow resnet50 time: 47.2ms (= 2362.1ms / 50, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 49.5ms (= 2473.0ms / 50, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.05 (= 49.5ms / 47.2ms)

OneFlow resnet50 time: 49.0ms (= 2447.8ms / 50, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 44.8ms (= 2238.9ms / 50, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 0.91 (= 44.8ms / 49.0ms)

OneFlow resnet50 time: 158.1ms (= 7903.3ms / 50, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 158.0ms (= 7900.3ms / 50, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.00 (= 158.0ms / 158.1ms)

OneFlow resnet50 time: 103.7ms (= 5184.4ms / 50, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 106.7ms (= 5337.5ms / 50, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.03 (= 106.7ms / 103.7ms)

OneFlow resnet50 time: 77.3ms (= 3867.2ms / 50, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 85.7ms (= 4285.3ms / 50, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.11 (= 85.7ms / 77.3ms)

OneFlow resnet50 time: 80.0ms (= 3998.6ms / 50, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 63.2ms (= 3158.8ms / 50, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 0.79 (= 63.2ms / 80.0ms)

OneFlow resnet50 time: 70.7ms (= 3534.1ms / 50, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 72.8ms (= 3638.9ms / 50, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.03 (= 72.8ms / 70.7ms)

[github-actions]

Speed stats:

GPU Name: GeForce GTX 1080 

OneFlow resnet50 time: 130.1ms (= 6504.0ms / 50, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 140.9ms (= 7047.0ms / 50, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.08 (= 140.9ms / 130.1ms)

OneFlow resnet50 time: 74.7ms (= 3733.2ms / 50, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 85.3ms (= 4265.2ms / 50, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.14 (= 85.3ms / 74.7ms)

OneFlow resnet50 time: 49.7ms (= 2485.5ms / 50, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 60.2ms (= 3010.3ms / 50, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.21 (= 60.2ms / 49.7ms)

OneFlow resnet50 time: 41.7ms (= 2082.8ms / 50, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 48.7ms (= 2432.7ms / 50, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.17 (= 48.7ms / 41.7ms)

OneFlow resnet50 time: 41.1ms (= 2054.6ms / 50, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 40.2ms (= 2011.4ms / 50, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 0.98 (= 40.2ms / 41.1ms)

OneFlow resnet50 time: 159.0ms (= 7949.5ms / 50, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 161.8ms (= 8091.7ms / 50, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.02 (= 161.8ms / 159.0ms)

OneFlow resnet50 time: 103.3ms (= 5166.4ms / 50, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 106.9ms (= 5346.3ms / 50, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.03 (= 106.9ms / 103.3ms)

OneFlow resnet50 time: 77.2ms (= 3859.2ms / 50, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 73.3ms (= 3663.5ms / 50, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 0.95 (= 73.3ms / 77.2ms)

OneFlow resnet50 time: 76.9ms (= 3845.9ms / 50, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 64.5ms (= 3224.0ms / 50, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 0.84 (= 64.5ms / 76.9ms)

OneFlow resnet50 time: 72.1ms (= 3605.3ms / 50, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 75.6ms (= 3779.6ms / 50, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.05 (= 75.6ms / 72.1ms)