Update hyperspectral calibration function

[nfahlgren]

Describe your changes
In evaluating pcv.hyperspectral.calibrate to address #843, a couple issues were found.

First, the original function was written assuming that the shape of the input dark and white reference images was (1, x, z) since this matched the data we worked with. @JorgeGtz noted that reference data provided in #843 had the shape (100, x, z), which caused the "memory" errors. To normalize the process for reference images with different numbers of line scans, the function is updated to average the lines (axis=0) and convert all reference inputs to the shape (1, x, z), which can be broadcast over the raw data cube shape.

Second, we realized that the calibration was done in two stages that were not connected, resulting in incorrectly calibrated output. The function was updated/simplified to do the calibration in a single vectorized operation.

Third, because calibrated values can be negative or greater than 1, we added a step to clip the values to the range 0-1. We confirmed that the resulting calibrated data are very similar to data calibrated by the manufacturer.

Last, we updated the function to utilize the _debug module.

Type of update
Is this a: Bug fix

Associated issues
Closes #843

[codecov]

Codecov Report

Merging #845 (45b8d17) into master (f1c0981) will not change coverage.
The diff coverage is 100.00%.

@@            Coverage Diff            @@
##            master      #845   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files          142       142           
  Lines         6223      6207   -16     
=========================================
- Hits          6223      6207   -16     

Flag	Coverage Δ
unittests	100.00% <100.00%> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
plantcv/plantcv/hyperspectral/calibrate.py	100.00% <100.00%> (ø)