chore: Initialize Docker setup

[tomas-gajarsky]

This pull request introduces a Docker-based environment setup for this CUDA-utilizing application, leveraging the power of Docker Compose to streamline development and execution on GPU-enhanced systems. By integrating a Dockerfile based on the CUDA 12.1 image and a corresponding Docker Compose configuration, we enable effortless building and running of the application in a GPU-accelerated environment. With this setup, executing the code in a predefined environment is as simple as running:

docker compose up

• Add Dockerfile based on CUDA 12.1 image
• Add compose config to build and run on GPU easily
• Add pyproject.toml with poetry dependencies
• Add poetry.lock with locked and verified dependency versions
• Add ignore files with common setup for Python repos
• Add debug logging and error catching to the torchrun main script
• Add build workflow for testing docker setup

[jiaweizzhao]

@tomas-gajarsky Thanks for providing the docker environment. Since we are still actively modifying the core training code, we are not planning docker wrap so far.