support local deployment & huggingface transformer

[pfliu-nlp]

It would be nice if users could use Promptware locally based on their local inference model (e.g., pipeline from huggingface)
(For example, users can replace openai's kernel with OPT.)

This would be very easy to implement using huggingface transformer & pipeline: https://huggingface.co/docs/transformers/pipeline_tutorial