Currently, we run server/wokers via ./bin/run which just calls server start-all.

This is fine & quick for dev mode, but for public machines it is suboptimal and causes confusion, like in #495 .

I propose to:

1. create a systemd job for server to just run: nix-shell --run "cabal v2-run gargantext -- server start --mode Prod --settings-path gargantext-settings.toml"
2. create a systemd job for workers: nix-shell --run "cabal v2-run gargantext -- worker run-all --settings-path gargantext-settings.toml"
3. create job templates e.g. under ./devops/systemd so that we have an idea how the job is constructed.

Having jobs in systemd is much better than some command being run in tmux:

• CPUQuota setting of systemd: we can limit CPU usage of workers, say, to 50% so that server slowness like in #495 is reduced.
• unified logs in journald
• automatic startup on boot, restarts on failure
• possibility to add dependencies, i.e. ggtx server/wokers won't start until postgres is started