Investigate possibilities of multiprocessing for Soledad Server

Some immediate possibilities are:

• Spawn many Soledad Server processes listening in different ports. In this case, requests should be routed to different processes/ports. It could be interesting to route the same user to the same process to improve the use of the per-process in-memory session caching. This could be done using python itself or something else (nignx, haproxy, etc). This option is only possible if there are no side-effects between requests, which is something we would have to make sure is the case.
• Spawn only one Soledad-Server process and use some multiprocessing library, like ampoule for example.