Need a fine-resolution bone_region mask

[jamesavery]

The coarse resolution bone_region mask was OK for gathering the statistics (2D-histogramming and optimizing the probability models).

However, it is unsufficiently accurate near the implant. For two reasons:

1. The coarse voxels make for an inaccurate interface to the implant.
2. The morphological closing (for including blood vessels in the region) also closed off the valleys from the implant threads: important regions for e.g. bone-to-implant contact that now can't be properly segmented!

Further: There are problems near the cylinder edge as well.

In conclusion: We need a better bone_region mask with as high resolution and accuracy as possible near the implant threads, and which covers the entire bone region.

[carljohnsen]

#37 Allowed for 0750 to run at 1x within reasonable time, thus producing a 1x mask. Some of the internal parameters (e.g. implant dilation) should however still be tuned in the future, to provide the tightest possible fits.