Improved pyroscan error handling (again)

[physycola]

The new implementation of pyroscan.py did not properly handle runs that failed to reshape correctly, which can sometimes happen with large batches of runs. I believe that this gives more robust error handling.

[FelixWattsYork]

Ah I see, When I change Pyroscan I ran into issues with how to handle creating a buffer is the first Pyro input file wasn't readable. This seems a good solution. My only request would be if you are able to create a test containing the original scenario that was failing. I wasn't aware of any good reason that you would have a pyroscan where the first item wasn't readable (hence the original error handling)

[physycola]

@FelixWattsYork , the failure state is unrelated to the pyroscan.py logic. In my case, it was related to the failure of cgyro.py to reshape the loaded gk_output data properly, likely due to a cluster-related IO failure. I am not really sure how to artificially manufacture such data, but more importantly, I don't think that falls under testing for pyroscan.py?