Entropy-Regularized Rank-Masked Policy Optimization for Test-Time Reinforcement Learning in Code Generation | TickerVault