fix: baseline always runs full rule sequence first, fix fill_nulls string mode

fix: fill_nulls mode on string columns (category/region), fix format crash

fix: real graders, real scores, /reset accepts empty body

Implement task scores and update scoring logic

Update environment.py

Update environment.py

Enhance score clamping for all task branches

Change initial previous score from 0.0 to 0.001

Update default score values in environment.py

DataClean OpenEnv v2.0 - Data Drift + Parallel Baseline