To properly visualize attack and release we calculate the input level out of
the gain reduction the output level threshold and ratio.
if (output_level >= threshold) { // no expansion
input_level = output_level
} else {
input_level = (gain_reduction - threshold * (1-ratio)) / ratio
}