Figure out how to encapsulate snaphu-py crashes/failures to not stop the entire workflow

[scottstanie]

If we are unwrapping many interferograms, we may not want a single failed snaphu-pu job to stop everything.

One possible reason- if we attempt multiple variants of unwrapping (ordering from easier/faster/more approximate to longer-running), we'd like to use earlier successes even if later jobs fail.

(this might be a better question to raise in snaphu-py for possible segfault handling changes)