instead of determining which component was clicked on, we're now looking fo rthe channel closest to the mouse y.
This allows larger channeloverlap, so there are fewer cases where the data get clipped.

just changed SAMPLES_PER_DATA_BLOCK back to 300, this probably requires
a real bug fix somewhere else, not sure why this is an issue.

still having an issue where the first sample on the left side of the screen is always drawn as 0, but its not a huge issue for now

20ms work very smoothly on a 1.2GHz core 2 duo laptop, so it's probably a safe value.