…finite alloc loop.

The situation arose when the condition

if ((dz = z[this_side] - z[this_side+1]) == 0.0f) continue;

was never met by a tiny bit. Comparing to 0.0 is never a good idea. So I made it

if (fabs((dz = z[this_side] - z[this_side+1])) < 1e-10) continue;

I get 6 grdcontour test failures but they don't seem related to this change (more due to the endless headache of slight different grids).

Fixes #8387