fix: use 0.01/0.99 bounds — 0.001/0.999 printed as 0.00/1.00 with 2dp format

fix: ensure no score boundary values (0.0/1.0) anywhere in grading pipeline

fix: ui logic, noise thresholds and identical code bypassing

feat: complete UI rebuild, isolated benchmarks, openenv schema fixes

fix: remove incompatible Python 3.12 pycache bytecode

initial commit