Currently bisect outputs an ungodly number of steps so that even with nesting, it is very hard to find out what is going wrong.

Furthermore, nesting is done in a way that may break the bisect when extensive retesting is required, since retesting nests incrementally.

Given that we are no longer waiting for builds asynchronously or testing multiple revisions at the same time, we can make better assumptions regarding the structure of the bisect and group steps appropriately.

We should at most have one step group per revision (allowing for repeats when retesting) and the grouped super step should at least indicate with the appropriate color if a failure happened within it.