[llbbl]

Closes #119

Summary

• Add deterministic, guarded benchmark data seeding for local performance runs.
• Add local-only k6 scenarios for HTTP message posting, message reads, and WebSocket fanout.
• Document reproducible baseline results and how to compare future runs.

Changes

Benchmark Harness

• app/Commands/SeedMessages.php: Add seed:messages with production and BENCHMARK_MODE safeguards, reserved users, and deterministic message data.
• app/Config/RateLimit.php: Relax rate limits only for non-production benchmark mode.
• bench/lib/auth.js: Share authenticated k6 session setup and endpoint configuration.
• bench/post-messages.js: Exercise 10-VU message posting for 30 seconds.
• bench/read-messages.js: Exercise 50-VU message reads for 30 seconds.
• bench/websocket-fanout.js: Measure 100-client WebSocket broadcast delivery.
• package.json: Add bench:post, bench:read, and bench:ws scripts.

Documentation and Tests

• bench/README.md: Document prerequisites, setup, commands, thresholds, and interpretation.
• docs/performance.md: Record baseline results against commit 12b314c with machine and runtime details.
• tests/Tests/Integration/SeedMessagesCommandTest.php: Verify guarded seeding, input validation, replacement behavior, and deterministic output.

Validation

• composer validate
• 166 unit tests / 497 assertions
• 10 integration tests / 56 assertions
• 1 WebSocket integration test / 8 assertions
• PHPStan
• CS check
• pnpm build
• 5 Playwright tests
• composer audit
• pnpm audit --prod
• k6 inspect for all three scenarios
• Actual post, read, and WebSocket benchmarks

Mutation testing remains manual-only through the workflow_dispatch workflow and was not added to automatic pull-request CI.