Fix bug about accessing model.config when model is None in pipeline

[regisss]

What does this PR do?

This PR fixes the issue mentioned in #297: when instantiating a pipeline with pipe = pipeline(task=my_task), the given model is None and we get the following error:

Traceback (most recent call last):
  File "test_yolos.py", line 3, in <module>
    classifier = pipeline(task="zero-shot-classification", accelerator="ort")
  File "/home/regis/HuggingFace/dev/tmp_venv/lib/python3.8/site-packages/optimum/pipelines.py", line 113, in pipeline
    load_tokenizer = type(model.config) in TOKENIZER_MAPPING or model.config.tokenizer_class is not None
AttributeError: 'NoneType' object has no attribute 'config'

With this PR, model.config is not accessed anymore to set load_tokenizer and load_feature_extractor. Tests have been added to automatically check this behaviour.

Also:

• the current version number is bumped,
• we were using the pipeline method from Transformers in tests/onnxruntime/test_modeling.py, this is changed for the pipeline method of Optimum.

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you make sure to update the documentation with your changes?
• Did you write any new necessary tests?

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint.

[philschmid]

LGTM

[philschmid]

LGTM. Nice that you reused the transformer's things of doing it. ✅