Remove deprecated old_benchmarking utilities

[Kovbo]

Summary

Remove the deprecated art.utils.old_benchmarking package instead of patching the unused display_image_grid helper reported in #691.

Changes

• Move the live reward_std_dev calculation helper into art.utils.trajectory_logging.
• Update Model.log() to use the current helper instead of importing from old_benchmarking.
• Delete the unused deprecated old_benchmarking modules, including display_image_grid.
• Add a focused assertion for the computed reward_std_dev value.

Test plan

• uv run pytest tests/unit/test_frontend_logging.py tests/unit/test_benchmarking_loader.py -q
• uvx ruff check src/art/model.py src/art/utils/trajectory_logging.py src/art/utils/benchmarking/aggregate_trajectories.py tests/unit/test_frontend_logging.py
• uvx ruff format --check src/art/model.py src/art/utils/trajectory_logging.py src/art/utils/benchmarking/aggregate_trajectories.py tests/unit/test_frontend_logging.py
• git diff --check