distinct_tokens argument to display_pca

[jonthegeek]

This combines with token_filter. The idea is to filter to "^train" (for example), and then just show one example each for "train", "training", "trained", "trains", etc (rather than a cloud). This should default to FALSE, but can be useful when constructing explanatory plots. Ideally we should choose the example at the "center" of that family, but I'm not sure that's important.