Summary

Benchmark targets in the Makefile pipe output through scripts/bench-tx-per-sec.awk to add a TX/sec column. This
can be done natively with Go's b.ReportMetric, which is more accurate for batch benchmarks and eliminates the
external script dependency.

Changes

• Add test.ReportTxPerSecond(b) helper in utils/test/benchmark.go
• Call it from all benchmarks across loadgen/, service/coordinator/, service/vc/, service/sidecar/,
  utils/deliverorderer/, and utils/serialization/
• For batch benchmarks (dependency graph, preparer): cap input to exactly b.N transactions so the metric is
  accurate
• Remove | awk -f scripts/bench-tx-per-sec.awk from all Makefile bench targets
• Delete scripts/bench-tx-per-sec.awk

Why

• Accuracy: The AWK script computes 1e9/ns_per_op, which gives ops/sec — not tx/sec for batch benchmarks where one
  "op" processes multiple transactions. b.ReportMetric reports the true throughput.
• Simplicity: No external script, no pipe — benchmark output includes tx/s natively.
• Consistency: All benchmarks report the same metric from a single helper.