Distributed CPU implementation

[dforero0896]

Methods missing to use distributed arrays. While the current reconstruction part works in a distributed fashion, the real priority for the reconstruction code would have to be distributing the catalog in a domain-aware form, as well as being able to read shifts in a similar manner.

• Create function that takes catalog array in main rank and distributes it.
  • This can be done via Scatter or distributing the relevant points to each process
  • The first is easier, and would allow reading different files by design. The second would avoid having checks in cic! and read_cic! (but I don;t feel comfortable removing them anyway) but would imply that it is necessary to know which process is the relevant one for a given point so data can be redistributed once in memory.