Feature/baseline fix

[jvmead]

Fixing a bug in the baselining:

• np.ptp is signed and requires np.abs around it to prevent occasional large offsets in the baseline for large signals
• dropping zeros from the list of sorted ranges prevents ADC clipping from generating a min range of 0 with offset of the full scale of the ADC as the baseline

[mjkramer]

LGTM with the proposed change: https://gist.github.com/mjkramer/ded775f35f7f5b91bea8757464d10465

[mjkramer]

Following up, new proposed change: Replace the zero ranges with nans only for the purpose of sorting. Use the true ranges (and means) everywhere else. This avoids returning nan while still achieving the original goal w.r.t. clipping. This seems to work (plots).

[jvmead]

I cherry picked your changes from Commit 927c659 linked above and broke it

[jvmead]

okay something weird is going on, I'll run another test locally