make sure to re-orient 2D mesh before meshing in 3D (fixes e.g. bug when 2D Frontal leaves badly oriented elements, that 3D frontal cannot handle)

Sigh++

No commit message

Previous version still used...

geometry. It introduces subtle bugs when reading high-order post-processing views.

IMHO we should have an explicit command to do this, and not modify the mesh without
warning just by reading a file

Test of metric as a shape measure. It uses the slot Rho instead of adding a new slot, define METRICSHAPEMEASURE in GmshDefines.h to activate !

prudently put between #if 0 #endif Those should allow
to produce better quality quad meshes and more robust
triangular meshes (random factor has disappeared) : 
Technically, the divide and conquer algo has been 
replaced by a bowyer-watson with robust predicates.

Cleaned code and window for high-order tools. Revamped fast curving (more robust and uses BL data when available). 

Call setVolumePositive during initial meshing and mesh optimization, as they can be performed independently

boundary layers in 3D are advancing (filtering intersecting layers using rtrees)
randomness of 2D meshes is now close to be ok (we still have problems with OCC ...)

IMPORTANT : 3D Delaunay Meshing now acts on connected volumes
	  separately, allowing to mesh regions that are not connected
	  topologically but that may intersect geometrically

Lloyd's algorithm needs to be applied before the mesh extrusion in order to keep identical starting and ending faces

Fopen deals with non-ASCII filename paths on Windows.

fix regression with extrusions of many surfaces: order faces by tag - this allows to minimize the number of passes necessary to resolve dependencies (otherwise 10 passes is not enough!)

natural convection added in the incompressible code