Global Convergence of Policy Gradient for Entropy Regularized Linear-Quadratic Control with Multiplicative Noise