Add new test /pgdb and one test implementation with expressJS

[MDA2AV]

Key details

• Postgres starts once before the profile loop, not per connection count. The
  100K-row seed loads in ~2 seconds via COPY.
• Only started when needed — if the framework doesn't have async-db in its
  meta.json tests, Postgres is never started.
• DATABASE_URL is only passed for the pgdb profile — other profiles
  (baseline, json, etc.) don't get it, so existing endpoints are unaffected.
• Validation uses a Docker network instead of --network host (because
  validate.sh uses port mapping). The framework and Postgres containers join
  httparena-validate-net so they can reach each other by container name.

[BennyFranciscus]

This is really well thought out 👏

A few things I like:

• No index on price — forcing a sequential scan is a great way to stress actual query throughput vs just index lookups. Makes it a real async I/O test rather than a "who has the fastest btree" contest.
• JSONB for tags, native BOOLEAN for active — cleaner than the SQLite version where you had to parse JSON strings and convert 0/1. Frameworks get to focus on what matters (async driver + pooling), not fighting type mismatches.
• The check_header semicolon fix — $expected_value;* accepting application/json; charset=utf-8 is a nice QoL improvement. Would have saved some pain on the drogon PR.
• Anti-cheat empty range check — simple but effective.
• engineScored: false in the composite leaderboard — correct call.

The benchmark/validate Postgres sidecar setup with the readiness loop is clean. Docker network for validate vs host networking for benchmark makes sense.

Express reference implementation with max: 16 pool size looks reasonable as a starting point.

One minor thought: the docs say "Recommended pool size: number of CPU cores or 16–32 connections" — might want to note that some async drivers (like asyncpg) perform better with smaller pools since each connection can multiplex, while thread-per-connection drivers need larger pools. But that's docs polish, not a blocker.

Looks good to merge! Excited to see how different frameworks stack up on this one.