Add the normalize and truncate parameters to feature_extraction to be compatible with Text Embeddings Inference (TEI)

[datavistics]

Is your feature request related to a problem? Please describe.
I would like to be able to use feature_extraction with full compatibility with Text Embeddings Inference: https://huggingface.github.io/text-embeddings-inference/#/Text%20Embeddings%20Inference/embed

Describe the solution you'd like
We should add the ability to put truncate and normalize in the post call

Describe alternatives you've considered
Ive tried workarounds, which involve overwriting the methods, but my friends made fun of my ugly code.

[Wauplin]

As explained in #1885 (comment), I prefer to put this feature request on hold until we have a proper parameter definition for inference. For now, let's close as "won't fix" and we get back to it once we have a more global approach to this.