Add support for scale_*(drop = FALSE)

I am having trouble getting plotly to respect a drop=FALSE argument for a geom_point() plot where I color by factor and only 1 of the factor levels is represented in the dataset. The code below generates 3 simple plots to demonstrate this

1. Simple iris plot, color by species, everything works properly
2. Same plot as 1, but only include data corresponding to a single species. The legend does not display any of the levels.
3. Same plot as 2, but adding the drop=FALSE argument to make the legend display all factor levels, even if they are not displayed in the dataset. The ggplot plot displays all levels in the legend, but the ggplotly version doesnt display any levels.

library(ggplot2)
library(plotly)

# plot all iris data
p <- ggplot(data=iris, aes(x=Petal.Width, y=Petal.Length, color=Species)) +
      geom_point()

print(p)
layout(ggplotly(p))

# just plot setosa species
setosa <- subset(iris, Species=="setosa")
p_setosa <- ggplot(data=setosa, aes(x=Petal.Width, y=Petal.Length, color=Species)) +
    geom_point()

print(p_setosa)
layout(ggplotly(p_setosa)
       , margin = list(r=100))

# setosa only, add drop=FALSE argument
p_setosa_nodrop <- ggplot(data=setosa, aes(x=Petal.Width, y=Petal.Length, color=Species)) +
  geom_point() +
  scale_color_discrete(drop=FALSE)

print(p_setosa_nodrop)
layout(ggplotly(p_setosa_nodrop)
       , margin = list(r=100))

Session info -------------------------------------------------------------
setting value
version R version 3.2.5 (2016-04-14)
system x86_64, mingw32
ui RStudio (0.99.486)
language (EN)
collate English_United States.1252
tz America/New_York
date 2016-04-19

Packages -----------------------------------------------------------------
package * version date source
assertthat 0.1 2013-12-06 CRAN (R 3.2.2)
base64enc 0.1-3 2015-07-28 CRAN (R 3.2.2)
colorspace 1.2-6 2015-03-11 CRAN (R 3.2.2)
DBI 0.3.1 2014-09-24 CRAN (R 3.2.2)
devtools 1.9.1 2015-09-11 CRAN (R 3.2.2)
digest 0.6.9 2016-01-08 CRAN (R 3.2.3)
dplyr 0.4.3 2015-09-01 CRAN (R 3.2.4)
ggplot2 * 2.1.0 2016-03-01 CRAN (R 3.2.4)
gridExtra 2.2.1 2016-02-29 CRAN (R 3.2.5)
gtable 0.2.0 2016-02-26 CRAN (R 3.2.3)
htmltools 0.3.5 2016-03-21 CRAN (R 3.2.5)
htmlwidgets 0.6 2016-02-25 CRAN (R 3.2.5)
httr 1.1.0 2016-01-28 CRAN (R 3.2.5)
jsonlite 0.9.19 2015-11-28 CRAN (R 3.2.5)
labeling 0.3 2014-08-23 CRAN (R 3.2.2)
magrittr 1.5 2014-11-22 CRAN (R 3.2.2)
memoise 0.2.1 2014-04-22 CRAN (R 3.2.2)
munsell 0.4.3 2016-02-13 CRAN (R 3.2.3)
plotly * 3.5.0 2016-04-19 Github (ropensci/plotly@43d882d)
plyr 1.8.3 2015-06-12 CRAN (R 3.2.2)
R6 2.1.2 2016-01-26 CRAN (R 3.2.3)
Rcpp 0.12.4 2016-03-26 CRAN (R 3.2.4)
scales 0.4.0 2016-02-26 CRAN (R 3.2.3)
tidyr 0.4.1 2016-02-05 CRAN (R 3.2.5)
viridis 0.3.4 2016-03-12 CRAN (R 3.2.4)
yaml 2.1.13 2014-06-12 CRAN (R 3.2.2)

Start by reproducing the three iris and setosa examples from the issue, comparing ggplot and ggplotly output for scale_color_discrete(drop=FALSE). Trace the ggplotly scale and legend conversion path; done means the setosa plot retains all factor levels in its legend when drop=FALSE is used, with the behavior covered by a regression test.