Optimal designs in three‐level cluster randomized trials with a binary outcome