Fix whisper for pipeline

[ArthurZucker]

What does this PR do?

After the merge of #19378 , the feature extractor does not work with the pipeline function. This PR is the same as #19385.

[HuggingFaceDocBuilderDev]

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

[patrickvonplaten]

Before we merge here, let's try to have the following tests working:

• Automatic pipeline tests for dummy whisper model (as discussed offline)
• 2 slow pipeline tests (one for speech recognition, one for speech translation here: https://github.com/huggingface/transformers/blob/10100979ed0594d4cfe1982cdfac9642a[…]/tests/pipelines/test_pipelines_automatic_speech_recognition.py

[ArthurZucker]

We were missing a _CHECKPOINT_FOR_DOC, so I added a warning when the tests are skipped.
It seems a little bit problematic as if it is unused, quality will fail (and in our case, I had to change the code to use it 😃 )
Other models that are not tested :
SpeechEncoderDecoderModel, Speech2TextForConditionalGeneration, which are also just missing the _CHECKPOINT_FOR_DOC(see here )

[patrickvonplaten]

Thanks for adding the tests!

[sgugger]

Thanks for fixing!