Fix OOM Killing Nomad Agent

[mpass99]

by increasing the memory reservation per runner.

Closes #676

This may not be the most stable solution as different environments (images) might require different amounts of memory. However, with increasing the memory reservation per runner only as many runners are allocated on the agent as it could handle right now (with the given server resources, and the allocation definition).

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 76.19%. Comparing base (e56b870) to head (16048c4).
Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #681   +/-   ##
=======================================
  Coverage   76.19%   76.19%           
=======================================
  Files          43       43           
  Lines        3592     3592           
=======================================
  Hits         2737     2737           
  Misses        625      625           
  Partials      230      230           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[MrSerth]

We have the memory oversubscription enabled anyway. Hence, we are still way below the actual memory footprint required when all jobs on an agent execute student code submissions.

[mpass99]

That's true, but so we might still have the OOM Killing behavior and also still missing visibility on it, preventing detection of such Nomad agent overloadings.