[Alexkkkkk]

Summary

• Capture detected mke2fs and debugfs versions in the benchmark report; require the relevant tool provenance for Linux artifact acceptance.
• Enforce bounded workload inputs: image <= 2 GiB, <= 256 paths, <= 256 MiB total extraction per iteration, <= 1 MiB range read per file, <= 100 measured repeats, and <= 20 warmups.
• Preflight file sizes before any temporary extraction, and have the artifact validator reject reports outside the declared bounds or with malformed paths.
• Document the limits and provenance fields. The benchmark remains read-only; no writer or cross-platform benchmark is claimed.

Validation

• pytest -q tests/test_benchmarks.py tests/test_benchmark_artifact.py — 27 passed.
• Ruff, Black, and git diff --check pass.
• Ran INCLUDE_10M=1 scripts/run_vfs_benchmarks.sh in a clean worktree; the artifact validator returned pass, source_worktree_dirty=false, native + pytsk3, with detected e2fsprogs 1.47.0 tool versions.

This is another bounded evidence tranche, not a claim that issue #390 is complete or that write workload, Windows/macOS provenance, or power-loss safety has been validated.

Refs #390