Implement the virtual fitting optimization metric

This issue follows https://github.com/OpenwaterHealth/OpenLIFU-python/issues/147, where only the minimal grid search approach is developed.

More specifically, we want to have a better definition for what a good position is:

1. Ray trace between each element and the target
   a. Get the material properties along the ray (checking only for sinus/non sinus region is sufficient, but we can also use a fully segmented head)
   b. Determine if the element is blocked
2. Calculate the number of blocked elements

Once the steering and number of blocked elements has been determined for each position, candidate positions whose steering is outside the valid steering range for the probe are excluded, and then the remaining points are filtered to those with the lowest decile of blocked elements. From those final candidates, the “optimized” position is then chosen as the one closest to the target.

Start by reviewing issue #147, which this work follows, and identify where the minimal grid search is implemented. Define how ray tracing, material checks, blocked-element counts, steering-range filtering, lowest-decile filtering, and target-distance selection fit together. Done means candidate positions are ranked by this optimization metric, but no files or tests are named in the issue.