fix: make sure to clean up any lingering processes

[Arqu]

Description

Occasionally there are lingering iroh/derper nodes running from tests (probably netsim failed runs) which might not be cleaned up properly (and might impact future runs on those boxes). This makes sure it kills anything remaining.

Notes & open questions

Change checklist

• Self-review.
• Documentation updates if relevant.
• Tests if relevant.

[flub]

this looks like a great reason to use cgroups in some capacity. So that an entire CI run is contained in some resource groups and you can make sure everything spawned in that groups is killed at the end.

systemd probably has facilities for this, though I'd also have to spend a while figuring those out. Anyway, only an idea to maybe use in the future. I'm kind of surprised this isn't part of the github runner already and that you have to deal with this here.

[Arqu]

It's a bit more involved I think. So the GH runner actually does do it's own stuff correctly. The issue here afaict is that the netsim depends on an external process that's already running on the box and manages coordination and spawning (which in this case is outside of the GH jurisdiction) on failed runs can leave loose nodes running. If I see it again I can check on the process tree to see exactly. Anyways this is a cheap way to do it for now.