feat(framework): openvino , openvino-model-server(ovms)

[KimSoungRyoul]

Feature request

This feature is likely to be a very low priority.
Also note that

• the triton openvino backend is currently being developed as an experimental version
  triton-inference-server/openvino_backend
• expectations for the performance of openvino optimizations in ovms in the future.

I'm not sure what level it would be efficient to support the openvino.

If someday I need this feature, it maybe more good idea to release an additional library like bentoml-openvino-extension.

1. support openvino IR model Runnable

   • https://docs.openvino.ai/2023.0/Supported_Model_Formats.html

2. support ovms bentoml[ovms] like a bentoml[tirton]

   • https://docs.openvino.ai/2023.0/ovms_what_is_openvino_model_server.html

3. docs to use intelpython with bentoml?