thresholding sox17

Sox17 has some bright spots on the surface of the gastruloids, but still dimmer than bright Bra or Sox2. The actual channel is hidden by the low dynamic range. Therefore, otsu only segments the spots, not the actual channel.

Solution: use 10 and 90 percentiles as min/max vals for clipping before thresholding.

Issue: this would always find a threshold even if the channel has no signal at all.