fix(end-to-end): stabilize fairies-v5 benchmark CI

[AztecBot]

Summary

• update client-flow benchmark step expectations to account for explicit mid-flow private-kernel resets
• let benchmark setup create multiple funded accounts and let sendTxs rotate sender accounts
• send the 32 standard build-block benchmark txs across four funded accounts to stay within the PXE sender tagging window

Investigation

AztecProtocol/aztec-packages#24223 was dequeued from merge-train/fairies-v5 because merge-group CI failed, not because of a merge conflict. The failed run showed:

• client_flows/account_deployments: expected 10 execution steps, got 11
• client_flows/amm: expected 14 execution steps, got 17
• bench_build_block: timed out after repeated PXE sender tagging-window errors: highest used index 21 was beyond the configured unfinalized window length 20

Verification

• env BENCHMARK_CONFIG=key_flows LOG_LEVEL=error BENCH_OUTPUT=bench-out/account-deployments.bench.json ./run_test.sh simple client_flows/account_deployments passed
• env BENCHMARK_CONFIG=key_flows LOG_LEVEL=error BENCH_OUTPUT=bench-out/amm.bench.json ./run_test.sh simple client_flows/amm passed
• env LOG_LEVEL=error BENCH_OUTPUT=bench-out/build-block.bench.json ./run_test.sh simple bench_build_block passed
• ./bootstrap.sh ci was attempted, but this branch's root bootstrap script exits immediately with Unknown command: ci before running tests

---

Created by claudebox · group: slackbot

[AztecBot]

Automatically closing this stale claudebox draft PR (no updates for 5+ days). Re-open if still needed.