Add "Run to Stable Result" option to new Anisotropy

[mdoube]

Describe the bug It is hard to guess what settings will result in a reasonable estimate of anisotropy for a given image. BoneJ1 included the option by default to keep sampling the image until the DA result stabilised using the mean and standard deviation of the preceding n DA estimates.

Expected behaviour Anisotropy should provide the user with the option (probably on by default) to keep going until a stable result is achieved. This may be engineered by adding directions and sampling them until a desired stability has been reached.

A new multithreading model may be needed that can tolerate new directions being added. Results can be dynamically added to a List, which is then used on each iteration to calculate DA stability.

See runToStableResult() in old Anisotropy to get an idea of how it's handled there.