The original proposal by Demsar includes a graphical indication of the critical distance itself, that we could add to our CDDs as well:

In some tests (such as Friedman-Nemenyi), being statistically significant simply means having ranks further apart than the critical distance. So this could also simplify the crossbar placement in these cases.
A first implementation idea would be to have a critical_distance argument that could optionally be passed instead of the sig_matrix argument. I'm busy for the next few weeks, but I intend to work on this issue in the future if people find this feature relevant enough and no one volunteers to implement this sooner :)