Potential improvement to statTally() -- bin splitting

[rpruim]

If the value of the test statistic is on the edge of a bin, then the shading added to the histogram can misrepresent the p-value a bit (since the data in a bin may not be uniformly distributed over the range of the bin).

The best, general fix is likely to split such bins. So all bins except bins containing the test statistic (possible also its additive inverse) would be the same size -- the exceptional cases would be narrower.

Can't decide whether this is worth the effort.