Support llama.cpp chat generation

[lbux]

Is your feature request related to a problem? Please describe.
Currently, the llama.cpp integration only supports create_completion() and not create_chat_completion(). Incorporating chat completion will allow for

• streaming
• function calling
• json constraining

Describe the solution you'd like
Proper implementation of create_chat_completion()

Describe alternatives you've considered
Ollama is built upon llama.cpp and supports Chat Completion, but there are reasons for not wanting to use Ollama if you already use llama.cpp

Additional context
The biggest reason for wanting to support this is json constraining. By allowing json constraining, output of an LLM can be structured in such a way that it can be used as input for evaluators that support (or can be modified to support) local llms through llama.cpp. Without constraining the output, parsing becomes difficult for models that are not as good at following directions.

[anakin87]

done in #723