DynUNet Torchscript Support

[ericspod]

Fixes #1245 .

Description

Adds Torchscript support to the DynUNet class. This passes the unit tests but may require further inspection by someone more familiar with this network to ensure it functions as expected still.

Status

WIP

Types of changes

• 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 --codeformat --coverage.
• Quick tests passed locally by running ./runtests.sh --quick.
• In-line docstrings updated.
• Documentation updated, tested make html command in the docs/ folder.

[Nic-Ma]

Thanks @ericspod for your quick update on TorchScript support.
@yiheng-wang-nv , could you please help review this PR? Let's make it happen in v0.4 release.

Thanks.

[yiheng-wang-nv]

Thanks @ericspod , I have checked the forward consistency for both train and val mode with our current version (I initialized these two versions of the networks with same weights), and the results are correct. I also run a training experiment under the Decathlon Dataset (task 4) and the performance is reasonable.

[ericspod]

Thanks guys. There's some cleanup I need to do and run the tests but then the PR should be ready to go.