Overhaul calibration process for VNIR/SWIR

[max-zilla]

Significant changes here, the result of several hackathons and collaborative effort to improve this extractor. Key changes:

• hyperspectral_workflow.sh script is not currently used anymore.
• python code is used to create empty netCDF file with appropriate headers. this empty file is then populated with the calibrated reflectance data once ready.
• calibrated reflectance data is written to file per-band, which should avoid the memory problems this extractor has had in the past and allow processing of all files.
• georeferencing of hyperspectral is corrected, not using terrautils still because it requires data from the .hdr file that terrautils is not prepared to locate in cleaning process.

So hyperspectral_calculation.py has the georeferencing code, calibrate.py creates empty netCDF, does calibration, calls georeference code, adds it all to headers.

THIS NEEDS MORE TESTING BEFORE I MERGE but should just be fixing bugs, not major changes. In separate pieces, all code has been tested.

[dlebauer]

This would be a good opportunity to address this previous issue https://github.com/terraref/reference-data/issues/83 and begin using CF standard names for the variables like reflectance http://cfconventions.org/Data/cf-standard-names/current/build/cf-standard-name-table.html