Improving the worthiness of the Henry problem as a benchmark for density‐dependent groundwater flow models