Allow to retry failed tasks automatically

[abner-math]

Allow to define in the infrabox.json specificiations how many tries Infrabox will perform before definitively failing a task. Create a new task status called "Flaky" for tasks which failed but got successful after one or more tries.

[ib-steffen]

The issue with such a feature is that many people tend to use it instead of looking into the actual root cause and fix it. So I tend to not add it, because it can make things actual worse and obviously add more load to the infrastructure. I'll leave it open for some further discussion though.