nx_js_h2h_bench.nx

nx_js_h2h_bench.nx -- HEAD-TO-HEAD: our sovereign VM/JIT vs V8 (node v22) on the SAME compute kernels, SAME machine. Kernels are the pure-compute subset our engine runs (no regex/prototypes/classes -> why we can't run full Octane yet). V8 numbers measured on THIS box (node, best-of-5, us) embedded as the ref so this gate prints the REAL cross-engine ratio -- NOT a vs-our-own-tree-walker number. EACH kernel's result is VERIFIED against V8's (stringified compare): a mismatch = FAIL (no ratio) and forces RED, so we can never report a ratio on a broken run. Kernel SIZES were chosen so BOTH engines complete; the limits our engine hit at larger sizes (array cap ~4096, call-count arena exhaustion ~fib27) are reported honestly. expect_exit: 0 license_tier: ORIGINAL