Update engine comparison benchmarks for Jint 4.9.0

[lahma]

Summary

Re-ran Jint.Benchmark/EngineComparison against the v4.9.0 tag with default BenchmarkDotNet settings, on the same hardware (Ryzen 9 5950X / Win11 / .NET 10.0.7 / BDN 0.15.8) used for the previous baseline. Refreshes:

• Last updated date → 2026-05-10
• Engine versions → Jint 4.9.0, NiL.JS 2.6.1722, YantraJS.Core 1.2.344
• Full results table

Notable v4.8.0 → v4.9.0 wins (Jint cold path, same machine)

Script	Time Δ	Alloc Δ
array-stress	−20.6%	−73.7%
dromaeo-object-array	−38.3%	−89.1%
dromaeo-object-array-modern	−43.6%	−89.1%
stopwatch	−8.8%	−45.8%
stopwatch-modern	−5.7%	−45.6%
dromaeo-core-eval	−9.7%	±0%
dromaeo-object-regexp (cached)	−13.0%	±0%

Driven by the perf PRs that landed in v4.9.0 — array allocation rework (#2451), member-expression inline cache (#2412), FunctionEnvironment pooling (#2413), recursion pool / JsString allocation (#2450), regex interpreter (#2388), and identifier-resolution fast paths.

Test plan

• Re-ran dotnet run -c Release --allCategories EngineComparison on the v4.9.0 tag (full default BDN, ~30 min)
• BDN reported 0 errors aside from the pre-existing Jurassic NA on modern syntax (already noted in the previous report)

🤖 Generated with Claude Code