ModuleManager: Reduce Buffer Per Worker to 256

I would argue that we can reduce the default buffer depth to 256. The reason is that in cases where only a long-running event is blocking the simulation, 256 slots per worker should mostly be enough (I see buffer depths of a few 100s of events being used - and this number scales with the number of worker threads).

In cases where the simulation is limited by IO, the buffer will fill very quickly anyway, but there is no added value of keeping more events in memory than necessary.

This will reduce the overall memory consumption.

Opinions?