Add script to benchmark tasks

[Tom0Brien]

This PR adds a script to benchmark both runtime and "control" performance of algorithms. This should help to compare various algorithms and make it easier to measure how changes impact runtime performance.

Usage

Run the benchmark on a specific task:

python hydrax/benchmark/run_benchmark.py --task Pendulum

TODO

• Add a more general configuration setup so that we can have task specific configs for both running examples and benchmarking
• General code clean up and simplify to make maintaining easy
• Add tests

Future plans

• Add script to perform automatic hyperparameter tuning using the benchmarking script (can actually probably use an algorithm from Evosax or optuna is good).

Example results

Benchmarking 5 controllers on task: Pendulum
Benchmarking PredictiveSampling on Pendulum...
Benchmarking MPPI on Pendulum...
Benchmarking CEM on Pendulum...
Benchmarking CMA_ES on Pendulum...
Benchmarking Sep_CMA_ES on Pendulum...
Benchmark complete in 66.43 seconds! Results saved to /home/tom/Projects/hydrax/hydrax/benchmark/results

Performance Summary:
Controller           Avg Cost        Final Cost      Avg Plan Time (s)    Realtime Rate  
------------------------------------------------------------------------------------------
PredictiveSampling   1.097615        0.000011        0.014874             0.95           
CEM                  1.974123        0.000000        0.012524             1.20           
Sep_CMA_ES           2.057362        0.000000        0.025399             0.68           
CMA_ES               3.726031        3.682149        0.032584             0.54           
MPPI                 3.818880        4.001694        0.012006             1.25