Remove the need for `worker_name` to simplify scaling

This issue has been migrated from [#8084](https://github.com/matrix-org/synapse/issues/8084).

---

**Description:**

The recent-ish changes to workers now recommend to set a unique `worker_name` per worker process. This makes scaling the number of workers quite a bit more complex than previously since now every instance needs a tailor made config file.

This is especially annoying when using things like docker-compose, swarm or kubernetes where spinning up multiple (identical) instances of a service is a built-in feature.

I have no concrete proposal on how to solve the problem(s) the worker name solves though. (AFAICS they're only really necessary for reverse mapping for federation senders and stream writers?)