Port improvements to color mapping functions from leaflet

[jcheng5]

• col_numeric(), col_bin(), col_quantile(), and col_factor() now support
  viridis colors. Just pass a palette name ("magma", "inferno", "plasma",
  or "viridis") as the palette argument.

• col_numeric(), col_bin(), col_quantile(), and col_factor() now have a
  reverse parameter, to apply color palettes in the opposite of their usual
  order (i.e. high-to-low instead of low-to-high).

• col_bin() and col_quantile() now take a right argument, which is passed
  to base::cut(); it indicates whether the bin/quantile intervals should be
  closed on the right (and open on the left), or vice versa.

• col_factor() now tries to avoid interpolating qualitative RColorBrewer
  palettes. Instead, it attempts to assign a palette color to each factor level.
  Interpolation will still be used if there are more factor levels than
  available colors, and a warning will be emitted in that case.

[jcheng5]

@hadley I've implemented your suggestions. There's one significant difference in behavior between my brewer_pal_colors and brewer_pal, that occurs when you request 2 colors from an RColorBrewer palette (such as when coloring a factor with only 2 levels). The RColorBrewer::brewer.pal call will return three colors; brewer_pal_colors will use the first and third, while brewer_pal will use the first and second.

Here's a demonstration of the differences:
http://rpubs.com/jcheng/small-palettes

I thought I'd feel more strongly about the leaflet version being superior, but now looking at that link, I'm not sure it's that big of a deal either way. I do think that for Sequential/Blues the leaflet version looks better, and for Qualitative/Accent the scales version looks better.

[hadley]

I don't feel particularly strongly, but I do think return the first two colours is more consistent that returning the first and the third.

[jcheng5]

@hadley I'm fine with that. Let me know if you see anything else I need to change--otherwise I think it's done.