feat: Automated Preset Docker Image Building

[ishaansehgal99]

This PR introduces a new standardized docker image for building all six llama models. Like so:

Build text completion model

# docker build \
#   --build-arg LLAMA_VERSION=llama-2-7b \
#   --build-arg SRC_DIR=pkg/presets/llama-2 \
#   -t llama-2-7b:latest .

Build chat completion model

# docker build \
#   --build-arg LLAMA_VERSION=llama-2-7b-chat \
#   --build-arg SRC_DIR=pkg/presets/llama-2-chat \
#   -t llama-2-7b-chat:latest .

This PR also defines github action for building and pushing llama and falcon models via az acr build service and git action