Document CI pinning and median-of-5 in benchmarks README

[intech]

Summary

Adds a short paragraph to `benchmarks/README.md` describing the CI wrapper landed in #15 (run-matrix-ci.sh: taskset -c 0 + median-of-5 + flat 5%/10% gates).

Secondary purpose

Triggers `benchmark.yaml` workflow end-to-end so we verify the post-#15 pinned baseline artifact against a pinned PR run. Expected: all fixtures within ±5% throughput, ±10% memory — previous runs had 6-7 false positives from unpinned runner noise.

🤖 Generated with Claude Code

[github-actions]

Benchmark: 1 regression(s)

Thresholds: throughput regression >5%, memory regression >10%. Runner pinned to CPU 0 via taskset. Current run on linux/x64, Node v22.22.2, captured 2026-04-20T17:09:46.760Z.
Baseline captured 2026-04-20T15:56:46.953Z on linux/x64, Node v22.22.2.

Summary: 1 regressed, 7 improved, 0 new, 12 unchanged.

Fixture	Baseline ops/s	PR ops/s	Δ ops	Baseline B/op	PR B/op	Δ mem	Status
SimpleMessage :: toBinary (pre-built, 19 B)	925,979	895,244	-3.3%	–	–	–	ok
ExportTraceRequest (100 spans) :: toBinary (pre-built, 32926 B)	1,137	1,249	+9.9%	–	–	–	improved
ExportMetricsRequest (50 series) :: toBinary (pre-built, 17696 B)	2,045	2,192	+7.2%	–	–	–	improved
ExportLogsRequest (100 records) :: toBinary (pre-built, 21319 B)	2,009	2,170	+8.0%	–	–	–	improved
K8sPodList (20 pods) :: toBinary (pre-built, 28900 B)	2,241	2,373	+5.9%	–	–	–	improved
GraphQLRequest :: toBinary (pre-built, 624 B)	186,483	176,832	-5.2%	–	–	–	REGRESSION
GraphQLResponse :: toBinary (pre-built, 1366 B)	245,970	244,497	-0.6%	–	–	–	ok
RpcRequest :: toBinary (pre-built, 501 B)	311,769	300,665	-3.6%	–	–	–	ok
RpcResponse :: toBinary (pre-built, 602 B)	458,123	445,231	-2.8%	–	–	–	ok
StressMessage (depth=8, width=200) :: toBinary (pre-built, 12868 B)	7,515	7,909	+5.2%	–	–	–	improved
SimpleMessage :: fromBinary (19 B)	1,076,345	1,034,031	-3.9%	–	–	–	ok
ExportTraceRequest (100 spans) :: fromBinary (32926 B)	615.3	631.0	+2.6%	–	–	–	ok
ExportMetricsRequest (50 series) :: fromBinary (17696 B)	1,173	1,200	+2.3%	–	–	–	ok
ExportLogsRequest (100 records) :: fromBinary (21319 B)	1,120	1,120	+0.0%	–	–	–	ok
K8sPodList (20 pods) :: fromBinary (28900 B)	1,304	1,418	+8.7%	–	–	–	improved
GraphQLRequest :: fromBinary (624 B)	297,876	306,930	+3.0%	–	–	–	ok
GraphQLResponse :: fromBinary (1366 B)	276,432	280,420	+1.4%	–	–	–	ok
RpcRequest :: fromBinary (501 B)	264,482	272,305	+3.0%	–	–	–	ok
RpcResponse :: fromBinary (602 B)	378,898	382,882	+1.1%	–	–	–	ok
StressMessage (depth=8, width=200) :: fromBinary (12868 B)	3,817	4,104	+7.5%	–	–	–	improved

Produced by benchmarks/scripts/compare-results.ts. Artifacts: bench-results-<pr> (current), bench-baseline-main (baseline).