nx_energy_model_test.nx

nx_energy_model_test.nx -- prove the state-independent energy model end to end, pure NishiLang, on THIS box (no PMU, no RAPL, on AC -- yet we still get a number). 1. MEASURE a charge-independent activity count (rdtsc cycles over a real workload). 2. ESTIMATE energy from the documented PRIOR coefficient (labeled as a prior). 3. CALIBRATE: given a real joule reading for a known activity, recover the coefficient and show energy(activity, calibrated) round-trips exactly -- this is the math a real RAPL/INA/discharge reading drives on a sensor-equipped node. 4. MONOTONIC: 2x the activity -> 2x the energy (sanity). The point proven: a real energy ESTIMATE with NO power sensor and IN ANY power state, because the input (cycles) is charge-independent. Known answer: exit 0.