* Fix eternal loop causing the application to hang
* Revert changes related to file well path introduced in 70bb22
* Use some iterations to find representative cells for computation of cell sizes
* Compute characteristic cell size based on active cells
* Compute cell face normal based on a suitable active cells
Using all cells as basis for face normal is fragile. Some models have highly distorted/twisted cells, but all active cells should be geometrically more stable.
The reason was a bit sloppy detection of whether the surface normals was
pointing outwards or inwards in a cell.
The cells can be turned inside out due to depth/Z or mapaxis conditions.
