Add a function to compare model error metrics as a function of number of replicates

[jds485]

For the supplement of the Regions paper, it would be good to show that using 20 replicates was sufficient. 20 is a small number, but the variability across the resamples was small.

Function should accept arguments for:

• set of error metrics computed from the current model replicates
• the number of bootstrap samples to create
• vector of the size of the bootstraps (up to a maximum of the current number available)

A plot function should also be created.

• x-axis is the size of the bootstrap samples from the previous function
• y-axis is the distribution (boxplot) of RMSE across the number of bootstrap samples