Use wave scenario in performance tests

[xsebek]

Is your feature request related to a problem? Please describe.

Our performance tests are out of date, we should measure realistic play scenarios.

Describe the solution you'd like

Let's use the Wave scenario (#1556) that has shown useful for measuring.

To truly stress the system and see how the number of running robots affects performance,
we should generate longer versions of the map. (e.g. 100, 300, 1000 robots).

Reusing the scenario description and just making longer maps should prove simpler
and more useful then writing the benchmarks by hand.

The advantage of using a scenario is that we could actually play it and see the difference.

Describe alternatives you've considered

We could write a new benchmark purely in Haskell, but that seems hard. (Terms can use TemplateHaskell though.)

We could also reuse the scenario except for the robots and place those dynamically like in other benchmarks.

Additional context

Ideally we would measure the overhead of drawing the map (terrain, entities and moving robots) as the base robot moves.

This could be achieved by serialising the visible area to Text in each tick, which could make it easier to inspect the benchmark.

[xsebek]

After reading the benchmarking code, I realised that the alternative is easier and makes it clearer what is tested.

Benchmarking map drawing and scenario condition evaluation should be separate benchmarks.