The SoftIndexStore still implements the old store SPI. We should convert this to the new SPI.

We should also evaluate to add some sort of segmentation support. As it stands currently, SIFS when segmented will create at least N * number of segments file descriptors and this could be 2 or 3 times that possible. Also the number of threads required is N * 6 as each store spawns a handful of threads.

Another issue is that now that we have PMEM support, we should really not offload to another thread as it is currently. It would be nice but not required (maybe another JIRA) to change how writes are done to do it in the invoking thread and do queuing for concurrent writes up to a limit of writes.