Tasks should carry separate timestamps for undo paths

[Zygmunt Krynicki]

For debugging tasks causality is important and it is currently seemingly not preserved when undo path is invoked. If we had separate timestamp for do-> and undo<- paths we would clearly see the sequence of events that happened on the system. I would also store the timestamp of an error that caused the undo to begin but I believe this is the "ready" timestamp of the do-> direction already.