nearWallFields does not produce correct sampling location

Summary

nearWallFields uses mappedPatchBase::facePoint to find a starting location which is on a tet-face when using the FACE_DIAG cell decomposition. Even on e.g. pitzDaily this might not find a point for all wall faces. Run e.g. nearWallFields with

type    nearWallFields;
fields
(
    (U UNear)
);

// Patches/groups to sample (regular expressions)
patches         (wall);

// Distance to sample
distance        0.00001;

the fourth face on the top-wall is not found correctly and it falls back to the cell centre.

2. the destination for tracking is calculated from the fall-back cell centre in above case instead of the wanted location on the wall face.

See the UNear on the top-wall. Above tiny distance (0.00001) should produce near-zero.