As part of #4117 we determined that the math.getOffset calculation is not correct and needs to be updated. It didn't produce any false-positives so we decided to update the code in another pr in order to get 4.8 out the door.

The algorithm should be as follows:

For targets with target size >= 24x24px, we should make sure we skip the (potentially expensive) offset calculation since the spacing exception isn't relevant. Target size === largest unobscured rect from math.splitRects. E.g. the result of calling dom.getTargetSize.

Otherwise, for undersized targets:

1. determine bounding rect of target and get the center (we'll call centerTarget). Bounding rect === the result of calling math.getBoundingRect for each rect in splitRects. (should be a new function)
2. determine the target's "neighbors" using grid (any other target that might be within 24px of any part of target's bounding rect is a neighbor)
3. for each neighbor:
   1. determine neighborSplitRects using math.splitRects(neighbor)
   2. for each neighborSplitRect:
      1. calculate distance from centerTarget to the nearest point in neighborSplitRect (may be 0 if centerTarget lies within neighborSplitRect)
   3. let neighborSpacing be the minimum neighborSplitRect distance
   4. if the neighbor is also undersized (target size < 24x24px):
      1. calculate centerNeighbor similarly to centerTarget using the neighbor's bounding box
      2. if the distance from centerTarget to centerNeighbor is < neighborSpacing, use that as the new neighborSpacing
4. return the minimum neighborSpacing from among all neighbors