Support for return_all_scores in pipeline

For `text_classification` pipeline, the parameter `return_all_scores=True` is needed to get all scores of all labels. Could this be integrated into the toolkit? (or potentially for newer version, the parameters is `top_k = n`).

Allowing to pass additional parameters needed for the pipeline would be great.

```
pipe = pipeline("text-classification", model='path/to/mode', tokenizer='path/to/tokenizer', return_all_scores=True)
```

Start at the toolkit's text_classification pipeline entry point and trace how pipeline arguments are handled. Exercise the supplied example with return_all_scores=True, or top_k=n for newer versions, and confirm that all label scores are returned; handling for additional pipeline parameters should also be explicit.

Written by the indexing model from the issue text.