[fx2trt] Better trt layer name (#67200)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/67200
We want to put more information on the tensorrt layer name. Mainly we want to be able to tell the original op that a TensorRT layer is mapped from.
The layer format is `[TensorRT Layer Type]-[Original Op Code]-[FX Node Name]`
```
Reformatting CopyNode for Input Tensor 0 to [FULLY_CONNECTED]-[acc_ops.linear]-[linear_1]: 0.0328ms
[FULLY_CONNECTED]-[acc_ops.linear]-[linear_1]: 0.027712ms
PWN([RELU]-[acc_ops.relu]-[relu_1]): 0.008672ms
```
Test Plan:
CI
```
buck run mode/dev-nosan -c python.package_style=inplace caffe2:fx2trt_example
```
Reviewed By: wushirong
Differential Revision: D31627274
fbshipit-source-id: 3dbb576caa63b922274541d2a306b4bd37e707c5