support free eager tensor in graph state dict

[BBuf]

在下面的pr进行了功能测试，现在带free eager tensor的onnx模型可以正常导出了。

Oneflow-Inc/oneflow_convert#89

[github-actions]

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9085/

[github-actions]

Speed stats:

GPU Name: GeForce GTX 1080 

❌ OneFlow resnet50 time: 130.0ms (= 13000.5ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 145.1ms (= 14508.2ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.12 (= 145.1ms / 130.0ms)

OneFlow resnet50 time: 77.7ms (= 7770.0ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 88.7ms (= 8870.8ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.14 (= 88.7ms / 77.7ms)

OneFlow resnet50 time: 49.6ms (= 9914.5ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 60.3ms (= 12061.9ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.22 (= 60.3ms / 49.6ms)

OneFlow resnet50 time: 36.5ms (= 7300.7ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 43.2ms (= 8641.4ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.18 (= 43.2ms / 36.5ms)

OneFlow resnet50 time: 32.4ms (= 6484.6ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 44.5ms (= 8896.8ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.37 (= 44.5ms / 32.4ms)

OneFlow swin dataloader time: 0.259s (= 51.752s / 200, num_workers=1)
PyTorch swin dataloader time: 0.157s (= 31.348s / 200, num_workers=1)
Relative speed: 0.606 (= 0.157s / 0.259s)

OneFlow swin dataloader time: 0.068s (= 13.620s / 200, num_workers=4)
PyTorch swin dataloader time: 0.040s (= 8.020s / 200, num_workers=4)
Relative speed: 0.589 (= 0.040s / 0.068s)

OneFlow swin dataloader time: 0.039s (= 7.737s / 200, num_workers=8)
PyTorch swin dataloader time: 0.022s (= 4.489s / 200, num_workers=8)
Relative speed: 0.580 (= 0.022s / 0.039s)

❌ OneFlow resnet50 time: 142.4ms (= 14242.2ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 165.1ms (= 16514.2ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.16 (= 165.1ms / 142.4ms)

OneFlow resnet50 time: 88.3ms (= 8832.9ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 110.6ms (= 11056.2ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.25 (= 110.6ms / 88.3ms)

OneFlow resnet50 time: 59.6ms (= 11925.6ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 89.1ms (= 17817.2ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.49 (= 89.1ms / 59.6ms)

OneFlow resnet50 time: 46.5ms (= 9305.6ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 70.3ms (= 14060.5ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.51 (= 70.3ms / 46.5ms)

OneFlow resnet50 time: 42.4ms (= 8470.3ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 75.7ms (= 15144.5ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.79 (= 75.7ms / 42.4ms)