Avoid building custom Caddy-based image for Docker deployment

[bkis]

Currently, we build a custom image for serving the client and proxying requests to the API based on caddy:x.x-alpine. The image build copied the client files from a previous build step. But as the final prod image is basically a vanilla Caddy image, we should instead do the following:

1. Build the client using a stock node image and store the distribution files locally (as opposed to in an image)
2. Run a stock Caddy image with two volumes (bind mounts) giving the container access to the Caddyfile and the static client files

[bkis]

Also, adjust the setup documentation accordingly!

[bkis]

This is trickier than expected, as our "one .env file provides all configuration"-approach isn't easily adaptable for e.g. a build.sh script that runs the client build in a container.

As there are more pressing issues right now and the technical as well as conceptual overhead of the current solution are negligible, this will be deferred.