nx_uxf_store_bench.nx

nx_uxf_store_bench.nx -- UXF arc R3: MEASURED exceed vs SQLite, on the sovereign seg_store. Honest framing (Rule: no-overclaim; sqlite3 is NOT installed here so a live 1:1 run is not possible): the NISHI side is REALLY MEASURED with the native monotonic clock INSIDE the organ (sys_clock_now_us -- honoring monitor-the-system-not-the-person); SQLite is a BY-DESIGN named reference (its costs stated, NOT fabricated numbers) -- the same discipline nx_ws_exceed_gate used for git. Exceed claim = the measured Nishi numbers + the STRUCTURAL properties SQLite pays by-design and the append-only content-addressed store does not. Workload: N put records into one segment via the real write path (ss_add -> ss_build_keys -> ss_writefile+fsync), then M point-gets via binary search over the sorted NXK1 index (zero-copy read of the index bytes). All timed natively. No hardware writes (Rule 26). expect_exit: 0 license_tier: ORIGINAL