fix: use 0.001/0.999 bounds — 1e-6 formats as 0.0000 in :.4f stdout which validator reads as 0.0

fix: eliminate all exact 0.0/1.0 paths, update reward_range in yaml and api to strict (0,1) bounds

fix: clamp task scores to strictly open (0, 1) interval for Phase 2 validator

Fix env loading for inference and add reward_space to openenv.yaml

fix: correct [START]/[STEP]/[END] log format to exact spec - lowercase key=value pairs, done=true/false, error=null, rewards as comma-separated floats with .4f precision

Update inference.py and openenv.yaml with improvements

Refactor GDPR Auditor: Full OpenEnv compliance & project cleanup

Initial commit