Test nginx as an alternative to apache2 as the Dashboard's production server

[ragesoss]

Current Behavior:

We currently use apache2 and Passenger to serve the Dashboard in production and staging (both for outreachdashboard.wmflabs.org and dashboard.wikiedu.org).

Occassionally, a production server ends up with a 'full queue' and stops working; this typically persists until apache (or the whole server) is restarted. Anecdotally, this may be caused by a sudden influx of users visiting at the same time (although it may, instead, have a root cause related to long-running web requests from background jobs that saturate the number of available connections).

Desired Behavior:

I'd like to be able to try out nginx (in a reversible way). Create a PR that can be deployed to staging (along with instructions for any server-side changes necessary) for doing this, so we can try it out and see if nginx lets us get around the capacity issues we've seen with apache.

[Bam92]

I wish I worked on this.

[ragesoss]

This was done as part of the infrastructure work to move the dashboard to use Terraform and Nomad for dashboard.wikiedu.org.

We still use Apache for outreachdashboard.wmflabs.org though, and we could definitely switch to nginx.