Re-enable functionality to pass through events which are too big for the GPU to process
When Allen encounters a slice of events which is too big to be processed in the GPU memory, it splits the slice and processes each split independently. This splitting continues, as necessary, until a single event remains. If a single event is too big to be processed, it should be passed through. This functionality used to be available but no longer works and should be brought back.
The event which is passed through must have a minimal SelReport and DecReport, so that it can be processed by Moore without requiring persistency hacks.
A counter (and associated histograms) should be added to monitor how often event slices get split, and how often individual events get passed through.