Write defragmented SION files when funneling output through Rank 0
Currently when funneling IO through Rank 0 we generate fragmented SION files as if IO had occurred in parallel. Theoretically everything could be written to 1 page in this case. This would not only speed up output but also generate considerably smaller SION files.
Is this worth implementing? It would definitely help for large datasets where the SION file could take up 40+ GiB.