These suites measure runtime latency, throughput, storage costs, and capacity limits. Run their commands from the repository root with the root Python environment.
Add load, stress, and capacity suites here, keeping each workload in its own directory with its setup and measurement contract. Put labeled-dataset quality evaluation in memory.