NIR spectra for mango fruit quality (DMC and cultivar/type/origin)

12011 samples, 306 spectral variables (wavelengths 285-1200 nm, 3 nm step).
Note: to keep the package under PyPI's size limit, the reflectance values
have been rounded to 4 decimal places and DM to 3 decimal places.

This is a multi-target dataset. Use the `target=` argument of
load_dataset()/open_dataset() to choose the y column(s), e.g.
    io.load_dataset('mango_nir')                        # default -> DM
    io.load_dataset('mango_nir', target='Cultivar')
    io.load_dataset('mango_nir', target=['DM','Cultivar','Type'])
    io.load_dataset('mango_nir', target='all')
A single target returns a 1-D y; multiple targets return a 2-D y.
String targets (Cultivar/Type/Region/Temp) are remapped to integer class
indices when selected alone.

Targets:
    DM       - Dry matter content (%), regression (default), range 9.46 ~ 24.58
    Cultivar - Mango cultivar, 10-class classification
               (Caly, KP, HG, Keitt, R2E2, 1243, LadyG, 1201, 4069, LadyJ)
    Type     - Ripeness, binary classification (Hard Green / Ripen)
    Region   - Growing region, binary classification (NT / QLD)
    Season   - Harvest season, 5 levels (1..5)
    Temp     - Storage temperature group (Low / Mid / High / No)

Task: regression (dry matter content) and/or classification.

Original publication:

    Anderson, N.T., Walsh, K.B., et al. (2020).
    Achieving robustness across season, location and cultivar for a NIRS
    model for intact mango fruit dry matter content.
    https://doi.org/10.1016/j.postharvbio.2020.111202
