[WIP]6577-Support-torch-trt-dynamo

[binliunls]

Fixes #6577 .

Description

PyTorch introduced the dynamo as a part to compile PyTorch models to different backends. And it also works for the Torch-TensorRT compiler and the Torch-TensorRT team is working on supporting it. Therefore, it would be better to have MONAI trt_export API also support this feature.

Currently the trt_export API only supports exporting models via torchscript models. Therefore a use_torchscript parameter will be added in this PR to the trt_export API to choose whether to use the torchscript as IR. Because the dynamo mode only supports to directly export models from pytorch models instead of the torchscript models.

Types of changes

• [x] Non-breaking change (fix or new feature that would not break existing functionality).
• [ ] Breaking change (fix or new feature that would cause existing functionality to change).
• [ ] New tests added to cover the changes.
• [ ] Integration tests passed locally by running ./runtests.sh -f -u --net --coverage.
• [ ] Quick tests passed locally by running ./runtests.sh --quick --unittests --disttests.
• [ ] In-line docstrings updated.
• [ ] Documentation updated, tested make html command in the docs/ folder.

[binliunls]

After discussing with the Torch-TensorRT team, currently there is not a way to serialize/export/save a dynamo model. So if we want to support the dynamo model, the best method as I know is to convert the model at the beginning of the inference. Since the conversion may take some overhead, the total inference latency may not necessary be optimized with this way. On the other hand, we can create a feature request for the torch-tensorrt team to add the support for the serialization.