Add support for Text Embeddings Inference (TEI).

[jpuerto]

• Update dependencies as previous version used is too old and fails with latest TEI docker images.
• Allows to use TEI by configuring an HTTP request value in model.

I had to limit the chunk-size to 510, not sure if that affects the quality of the embeddings. There should be a way to increase this size in the TEI service because ATM it is using 512 as limit and I did not found how to change it.