Run osc without binlogging and with no-swap-tables

Being able to run OSC without binlogging and without swap-tables would enable us to run OSC in parallel on all slaves then, when done, run the table swap on the master so they all get swapped at the right time (relative to the writes on the master).

Running OSC in parallel, locally, on every slave speeds it up and relieves the replication thread of unnecessary work, also the network doesn't have to move huge amounts of data (size of table X number of slaves) that can go in the petabyte range...