Production deployment

[skivdal]

Depends on #6.

The application should be possible to deploy quite easily. To achieve this, we should implement the following tasks:

• Dockerfiles for frontend and backend
  • Backend dockerfile should be built with a docker image containing CUDA, as some worker tasks depends on GPU
• Backend docker container needs an entrypoint script, selecting if it is a worker or server
• nginx reverse proxy config
• docker-compose with all services we depend on, including frontend and backend
• Domain name and certbot configuration for HTTPS.

For the concrete deployment, we should also contact NREC for advice regarding dynamic allocation of GPUs and object storage (s3 compatible) resources.