Weight random test content with word frequency data

[max-niederman]

Currently, when elements are selected from a language file, each word is given equal weight; however, in reality, some words are much more common than others. Weighting each word by how often it's used would result in more realistic test content.

This shouldn't be too difficult to implement since rand already has a method to choose random elements with weights, but it will require changing the format of the language file.