Add Optimum CLI backbone

[fxmarty]

This PR adds the backbone for an Optimum CLI, largely inspired from Transformers. Two commands are for now available:

optimum-cli export
optimum-cli env

and the export can handle several sub parsers, with the only one supported currently being

optimum-cli export onnx [ARGS]

that just maps to python -m optimum.exporters.onnx [ARGS].

In the future, this CLI could support things like

optimum-cli onnxruntime optimize -Ox --model model_name_or_path
optimum-cli onnxruntime quantize --dynamic --arch avx2 --model model_name_or_path
optimum-cli habana ???? (if relevant)
optimum-cli openvino quantize ???? (if relevant)
optimum-cli intel-neural-compressor quantize ???? (if relevant)

Fixes #188

[HuggingFaceDocBuilderDev]

The documentation is not available anymore as the PR was closed or merged.

[michaelbenayoun]

For me that looks very interesting!
Especially because it looks easily extendible and do not make much assumptions on the underlying features.

[regisss]

LGTM!

[JingyaHuang]

LGTM