Modularize real-world benchmark modules

[yvette-carlisle]

Summary

• convert the real-world job benchmark binary to a normal directory-style module tree
• split scoreboard, markdown, operational, and external adapter validation code into owned Rust submodules
• update current specs/runbook references for the new benchmark entrypoint

Validation

• cargo make fmt-rust && cargo make check after each segment
• final local check: 403 passed, 92 skipped

Batch size

• 55 files changed, 1706 insertions, 1569 deletions