[WIP] 8134 Add unit test for realtime inference

[Nic-Ma]

Fixes #8134 .

Description

This PR added unit test to cover the realtime inference with bundles.
And updated BundleWorkflow to support cyclically calling the run function with all components instantiated.

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 -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.

[ericspod]

Hi @Nic-Ma, we haven't had time to discuss the ideas around realtime/streaming inference yet due to MICCAI, but I feel we should look into doing this with specialised versions of datasets and transforms to represent streams of incoming and outgoing items. This would implement streaming without modifying existing workflows if we can get it performant.

[Nic-Ma]

Hi @ericspod ,

Yes, I totally agree with you, I was thinking both the streaming way and repeatedly calling way.
Actually, for these 2 calling methods:
(1) pending IterableDataset in a separate thread for the real-time inference input streaming data.
(2) Repeatedly call BundleWorkflow.run() with replaced input data, just what I made the simple test case in this PR.

This PR is just a draft implementation for discussion, could you please help also write draft code for your idea and let's discuss later?
When you design the logic, please consider below typical use cases:
(1) Triton backend: https://github.com/triton-inference-server/python_backend?tab=readme-ov-file#usage
(2) MONAI deploy: https://github.com/Project-MONAI/monai-deploy-app-sdk/blob/main/monai/deploy/operators/monai_bundle_inference_operator.py#L277
(3) Holoscan inference: https://github.com/nvidia-holoscan/holohub/tree/main/tutorials/dicom_to_usd_with_monai_and_holoscan#monai-deploy--holoscan-pipeline

Thanks in advance.

[ericspod]

This may be a good addition regardless of the intent here.

[ericspod]

This prevents keeping large objects created by the expression evaluation which aren't used anymore, so might be an optimisation in some cases.