Nuke inference plugin

For inference of a Image to Image network in Nuke - RMTC CAT building is in the pipeline - however using CAT to execute a network does not allow RMTC to directly track the inference result, this would still be required to be populated, in addition CAT files operate on specifically wrapped Torchscript files.

Creating an RMTC inference plugin that indexes off solutions rather than weights or models will permit flexible 'best run' interfacing, multiple model type support and allow the inferences to be correctly recorded to RMTC as they occur.

Start by reviewing how Nuke CAT executes wrapped TorchScript files and how RMTC currently records inference results. Define the plugin’s solution-based indexing and confirm that different model types can be executed while each inference is correctly populated in RMTC.

Written by the indexing model from the issue text.