Logarithmic-Regret Quantum Learning Algorithms for Zero-Sum Games