Keynote 2 benchmark updates & refinements

[bradleyshep]

This PR builds on #4978

Description of Changes

Methodology refresh of templates/keynote-2/.

• Pool size fix. Removed the poolMax = 1000 override on the pg-flavored RPC servers (postgres, cockroach, supabase). All now use node-pg's default of 10, matching bun-server.ts.
• Convex retry loop removed. The 32-retry client-side loop in connectors/convex.ts was masking conflict-storms with a death spiral.
• Convex package bumped from ^1.29 to ^1.37.
• Hidden Postgres tuning removed from docker-compose.yml. Local pg keeps only the explicit default_transaction_isolation = 'serializable' override.
• Per-second time-series sampling added to core/runner.ts. Each run emits a timeSeries array for warmup/decay/collapse charts.
• Bench CLI accepts --alpha 0,1.5 (CSV), --runs N, and --prep-between-alphas, writing one JSON per (connector, alpha, run) tuple.
• Helper scripts added: start-bench.sh, stop-bench.sh, check-bench.sh, bench-stats.py, plot-bench.py. Paths resolve relative to the script file.
• README updated with current measurements, two PlanetScale rows (PS-2560 single-node EBS and M-15360 Metal HA), and isolation-level methodology notes.
• Removed unused Docker files

API and ABI breaking changes

No engine API/ABI changes. Internal shape changes for the keynote template:

• BenchOptions.alpha: number → alphas: number[]. New fields: runs, prepBetweenAlphas.
• RunResult gained a required timeSeries field.
• Output JSON filename is now test-1-<connector>-a<alpha>-<timestamp>.json (glob test-1-*.json still matches).

Expected complexity level and risk

3. Contained to templates/keynote-2/ but cross-cutting within it (runner, CLI, RPC servers, docker-compose, scripts, README).

Testing

Testing

Reviewer smoke test:

pnpm run prep
pnpm run bench --alpha 0,1.5 --connectors postgres_rpc --seconds 30 --runs 1
ls runs/test-1-postgres_rpc-a*.json   # expect 2 files

• pnpm build passes
• pnpm run bench --help lists the new flags
• Reviewer: smoke sweep above produces 2 JSONs (one per alpha)
• Reviewer: confirm --prep-between-alphas actually re-preps between alphas
• Reviewer: confirm pnpm run demo still works