updates to aggregateColor()

[dylanbeaudette]

Performance and accuracy updates:

1. when snapping to k colors, clustering should be applied to unique colors
2. clustering should be performed on delta-E00 distances
3. use new pamonce=5 argument to cluster::pam for better performance
4. add option for weighting by site-level attributes
5. input column names should be checked for better error reporting
6. output should include sigma from rgb2munsell back-transform, or re-generate a representative color based on the closes Munsell chip
7. standardization in pam(LAB, ..., stand=TRUE) is not correct

An example for item 6:

library(aqp)

# two slightly different colors, same chip
cols <- t(col2rgb(c('#5F5345', '#554636'))) / 255
rgb2munsell(cols)

hue	value	chroma	sigma
10YR	3	2	0.0420
10YR	3	2	0.0016

[dylanbeaudette]

Initial testing of no. 1,2,3,7 above.

The results are reasonable and 40x speed improvement. Using stand=TRUE was clearly a mistake.

[dylanbeaudette]

Added testing code to aqp/misc/test_suite/aggregateColor-dev.R.