Recovery manager #96
Description
When functions execute, they create backups of the output they produce (i.e., the input(s) to the next function(s)) when they are finished executing (i.e., checkpointing). These backups should be used by a recovery manager in case of a failure of a subsequent function.
There should be some coordination with the recovery manager regarding when functions start, finish and fail.
See #85 for how these checkpoints are created and can be found.