A Penalized Poisson Likelihood Approach to High-Dimensional Semi-Parametric Inference for Doubly-Stochastic Point Processes