Tuning Sequential Monte Carlo Samplers via Greedy Incremental Divergence Minimization