Use number of bins instead of quantization interval in mlos_bench tunables

Use number of bins instead of quantization interval in mlos_bench tunables #3753