Slightly decrease size of webserver image by excluding unnecessary dependencies.

This release moves the project into a more streamlined direction.

It moves away from the attempt to detect languages and do the same thing for all kinds of different languages. An instance of lingolift is supposed to support only one or few similar languages. This release reflects that approach in defining the spaCy models that can be used for sentence analysis on an image level. More changes have to and will be made to the cloud infrastructure in the future.

Additionally, self-hosting this application becomes more accessible via a set of language-specific images hosted in Dockerhub (e.g. tobiaswaslowski/lingolift-webserver-ru for Russian). These images provide a Flask webserver that emulates the API Gateway infrastructure that underlies my hosted services on the Streamlit Community Cloud and Telegram.

• Put Lambdas behind API Gateway, enabling drastic speedups with the aiohttp client

• Handle requests via Lambdas instead of the REST webserver
• Create CI/CD setup for automated deployments