Fixes #4145: Add support for huggingface models

[vga91]

Test some tomasonjo models for text2cypher with LocalAI

We have to deploy the model (with LocalAI in this case), as we cannot use the public remote HF Inference API, see e.g. here,

For instance, by using inference we receive:

.. Request
$ curl https://api-inference.huggingface.co/models/tomasonjo/text2cypher-codestral-16bit \
        -d '{"inputs": "What color is the sky?"}' \
        -H 'Content-Type: application/json' \
        -H "Authorization: Bearer <API_TOKEN>"

.. response 

{"error":"The model tomasonjo/text2cypher-codestral-16bit is too large to be loaded automatically (44GB > 10GB). Please use Spaces (https://huggingface.co/spaces) or Inference Endpoints (https://huggingface.co/inference-endpoints)."}%   

..
.. Request
$ curl https://api-inference.huggingface.co/models/tomasonjo/text2cypher-demo-6bit-gguf \
        -d '{"inputs": [{"role":"user", "content":"Retrieve distinct values of the title and the comments from Article where title is not The Gervais-Neveu-Felder"}]}' \
        -H 'Content-Type: application/json' \
        -H "Authorization: Bearer <API_TOKEN>"

.. response 
{"error":"Task not found for this model"}