nx_native_micro_bench.nx

nx_native_micro_bench.nx -- minimal NATIVE x86_64 Nishi micro-bench. Goal: measure a tight integer-loop workload running NATIVE x86_64 (no qemu), output the wall-clock to stdout, return exit-code = 0. Apples-to-apples comparator vs the equivalent C loop run on the same hardware. Surfaces the substrate's REAL per-op cost when freed from the qemu interpreter (~100-1000x overhead). Workload: sum integers 0..N-1 with a register-only loop. Trivial math; the comparison is "how fast can Nishi emit tight loops on x86_64 vs gcc -O2 on the same loop."