Global Convergence of Policy Gradient for Linear-Quadratic Mean-Field Control/Game in Continuous Time