Handle incomplete benchmark results safely

[Daniiiil1]

Summary

• avoid masking a missing benchmark result field with an UnboundLocalError
• use the nested result mapping consistently while parsing
• report an unknown timestamp when the timestamp itself is missing
• add a regression test proving invalid entries are skipped and later valid results still load

Testing

• pytest benchmarks/other/ros2benchmark/test/test_benchmark.py -q (1 passed)
• flake8 benchmarks/other/ros2benchmark/test/test_benchmark.py
• flake8 benchmarks/other/ros2benchmark/ros2benchmark/verb/__init__.py --select=E9,F63,F7,F82
• python -m compileall
• git diff --check

Signed-off-by: Daniil Mordanov mordanovdania@gmail.com