Calculate confidence intervals based on race means

First calculate the mean per race, then use that to calculate the confidence interval. I feel it should be more representative of the actual typing performance, because some texts are really short and may have been typed several times.

Should investigate the mathematical theory behind this. I know there is a trick that can be done with variances, but I need to dig into it some more.

No file or test is named in the issue. Start by locating the existing race mean and confidence-interval calculations, then review the statistical theory for combining variances across races. Done means confidence intervals are calculated from per-race means and account for repeated or very short texts.

Written by the indexing model from the issue text.