Note that with the library superb (which I am the maintainer), you do not need to worry about aggregating the scores. Simply use

library(ggplot2)
library(superb)

superb(MeanDecreaseGini ~ Feature, data) +
    coord_flip() +
    theme_classic() +
    labs(
      x     = "Feature",
      y     = "Importance",
      title = "Feature Importance")

from which you get the plot with 95% confidence interval (default), or adding the option errorbar = "SE" to get standard error: