Allow re-trying jobs when appropriate

[atruskie]

Some jobs timeout or fail because resource load is high or external services are down.

For some failures we should implement a retry policy. SideKiq does this with automatic backoff policies.

Example of intentional failures that should not be retried: recording too short when harvesting exception, recording empty (0 bytes) exception, etc...

Example of failures that should be retried: time outs from audio tools, ...

[atruskie]

https://github.com/lantins/resque-retry