Performance testing at different numbers of channels, event rates, and chunk sizes

[cboulay]

I'm particularly concerned about the performance of my sliding_win_oneaxis implementation for sparse arrays.