Run fresh baselines

For each benchmark, we'd want:
- static actions
- random dynamic
- tuned RL
- any known solution policies (e.g. CSA)
- optimal solution if it exists

Start by identifying every benchmark covered by DACBench and how each benchmark is currently run. For each one, establish runs for static actions, random dynamic, tuned RL, known solution policies such as CSA, and an optimal solution where available. The work is done when fresh baseline results exist for every benchmark and the run details are recorded.

Written by the indexing model from the issue text.