Verify operator IO data types with typeguard

[gigony]

• Verify the data type of Operator input/output by using [typeguard](.
• Fix existing IO type errors in DICOMDataLoaderOperator and DICOMSeriesSelectorOperator\

Resolves #59

The cyclic graph seems to be handled by default when using NetworkX (but may need to check explicitly in the future).

     79             # Set source input for a label if op is a root node and (<data type>, <storage type>) == (DataPath, IOType.DISK)
     80             is_root = g.is_root(op)
---> 81             if is_root:
     82                 input_op_info = op.op_info
     83                 input_labels = input_op_info.get_labels(IO.INPUT)

~/miniconda3/envs/monai/lib/python3.6/site-packages/networkx/algorithms/dag.py in topological_sort(G)
    195     if indegree_map:
    196         raise nx.NetworkXUnfeasible(
--> 197             "Graph contains a cycle or graph changed " "during iteration"
    198         )
    199 

NetworkXUnfeasible: Graph contains a cycle or graph changed during iteration

[MMelQin]

Existing example apps that showcases DICOM input parsing all use the DICOM Loader and the Series Selector in tandem, so the change from Study to List[Study] does not impact existing sample in the repo.

[gigony]

Existing example apps that showcases DICOM input parsing all use the DICOM Loader and the Series Selector in tandem, so the change from Study to List[Study] does not impact existing sample in the repo.

Thanks Ming!

With this PR, it actually checks types in runtime so without changing Data type to correct type, it caused runtime error.

E.g.,

 @input("dicom_files", DataPath, IOType.DISK)
-@output("dicom_study_list", List[DICOMStudy], IOType.IN_MEMORY)
+@output("dicom_study_list", DICOMStudy, IOType.IN_MEMORY)
 @env(pip_packages=["pydicom >= 1.4.2"])
 class DICOMDataLoaderOperator(Operator):

$ python examples/apps/ai_spleen_seg_app/app.py -i dcm/ -o output -m model.pt

Traceback (most recent call last):
  File "examples/apps/ai_spleen_seg_app/app.py", line 79, in <module>
    app_instance.run()
  File "examples/apps/ai_spleen_seg_app/app.py", line 43, in run
    super().run()
  File "/home/gbae/miniconda3/envs/monai/lib/python3.6/site-packages/monai/deploy/core/application.py", line 322, in run
    executor.run()
  File "/home/gbae/miniconda3/envs/monai/lib/python3.6/site-packages/monai/deploy/core/executors/single_process_executor.py", line 120, in run
    op.compute(op_exec_context.input_context, op_exec_context.output_context, op_exec_context)
  File "/home/gbae/miniconda3/envs/monai/lib/python3.6/site-packages/monai/deploy/operators/dicom_data_loader_operator.py", line 56, in compute
    output.set(dicom_study_list)
  File "/home/gbae/miniconda3/envs/monai/lib/python3.6/site-packages/monai/deploy/core/io_context.py", line 115, in set
    f"The data type of '{label}' in the {self._io_kind} of '{self._op}' is {data_type}, but the value"
monai.deploy.exceptions.IOMappingError: The data type of 'dicom_study_list' in the output of '<monai.deploy.operators.dicom_data_loader_operator.DICOMDataLoaderOperator object at 0x7f31929ae7b8>' is <class 'monai.deploy.core.domain.dicom_study.DICOMStudy'>, but the value to set is the data type of <class 'list'>.

monai-deploy-app-sdk/monai/deploy/operators/dicom_data_loader_operator.py

Line 103 in c7d4868

return list(study_dict.values())